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 37#include <sys/cdefs.h> 38#include "opt_capsicum.h" 39#include "opt_ktrace.h" 40 41#include <sys/param.h> 42#include <sys/systm.h> 43#ifdef COMPAT_FREEBSD11 44#include <sys/abi_compat.h> 45#endif 46#include <sys/bio.h> 47#include <sys/buf.h> 48#include <sys/capsicum.h> 49#include <sys/disk.h> 50#include <sys/malloc.h> 51#include <sys/mount.h> 52#include <sys/mutex.h> 53#include <sys/sysproto.h> 54#include <sys/namei.h> 55#include <sys/filedesc.h> 56#include <sys/kernel.h> 57#include <sys/fcntl.h> 58#include <sys/file.h> 59#include <sys/filio.h> 60#include <sys/limits.h> 61#include <sys/linker.h> 62#include <sys/rwlock.h> 63#include <sys/sdt.h> 64#include <sys/stat.h> 65#include <sys/sx.h> 66#include <sys/unistd.h> 67#include <sys/vnode.h> 68#include <sys/priv.h> 69#include <sys/proc.h> 70#include <sys/dirent.h> 71#include <sys/jail.h> 72#include <sys/syscallsubr.h> 73#include <sys/sysctl.h> 74#ifdef KTRACE 75#include <sys/ktrace.h> 76#endif 77 78#include <machine/stdarg.h> 79 80#include <security/audit/audit.h> 81#include <security/mac/mac_framework.h> 82 83#include <vm/vm.h> 84#include <vm/vm_object.h> 85#include <vm/vm_page.h> 86#include <vm/vnode_pager.h> 87#include <vm/uma.h> 88 89#include <fs/devfs/devfs.h> 90 91MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information"); 92 93static int kern_chflagsat(struct thread *td, int fd, const char *path, 94 enum uio_seg pathseg, u_long flags, int atflag); 95static int setfflags(struct thread *td, struct vnode *, u_long); 96static int getutimes(const struct timeval *, enum uio_seg, struct timespec *); 97static int getutimens(const struct timespec *, enum uio_seg, 98 struct timespec *, int *); 99static int setutimes(struct thread *td, struct vnode *, 100 const struct timespec *, int, int); 101static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred, 102 struct thread *td); 103static int kern_fhlinkat(struct thread *td, int fd, const char *path, 104 enum uio_seg pathseg, fhandle_t *fhp); 105static int kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg, 106 size_t count, struct thread *td); 107static int kern_linkat_vp(struct thread *td, struct vnode *vp, int fd, 108 const char *path, enum uio_seg segflag); 109 110uint64_t 111at2cnpflags(u_int at_flags, u_int mask) 112{ 113 uint64_t res; 114 115 MPASS((at_flags & (AT_SYMLINK_FOLLOW | AT_SYMLINK_NOFOLLOW)) != 116 (AT_SYMLINK_FOLLOW | AT_SYMLINK_NOFOLLOW)); 117 118 res = 0; 119 at_flags &= mask; 120 if ((at_flags & AT_RESOLVE_BENEATH) != 0) 121 res |= RBENEATH; 122 if ((at_flags & AT_SYMLINK_FOLLOW) != 0) 123 res |= FOLLOW; 124 /* NOFOLLOW is pseudo flag */ 125 if ((mask & AT_SYMLINK_NOFOLLOW) != 0) { 126 res |= (at_flags & AT_SYMLINK_NOFOLLOW) != 0 ? NOFOLLOW : 127 FOLLOW; 128 } 129 if ((mask & AT_EMPTY_PATH) != 0 && (at_flags & AT_EMPTY_PATH) != 0) 130 res |= EMPTYPATH; 131 return (res); 132} 133 134int 135kern_sync(struct thread *td) 136{ 137 struct mount *mp, *nmp; 138 int save; 139 140 mtx_lock(&mountlist_mtx); 141 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) { 142 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) { 143 nmp = TAILQ_NEXT(mp, mnt_list); 144 continue; 145 } 146 if ((mp->mnt_flag & MNT_RDONLY) == 0 && 147 vn_start_write(NULL, &mp, V_NOWAIT) == 0) { 148 save = curthread_pflags_set(TDP_SYNCIO); 149 vfs_periodic(mp, MNT_NOWAIT); 150 VFS_SYNC(mp, MNT_NOWAIT); 151 curthread_pflags_restore(save); 152 vn_finished_write(mp); 153 } 154 mtx_lock(&mountlist_mtx); 155 nmp = TAILQ_NEXT(mp, mnt_list); 156 vfs_unbusy(mp); 157 } 158 mtx_unlock(&mountlist_mtx); 159 return (0); 160} 161 162/* 163 * Sync each mounted filesystem. 164 */ 165#ifndef _SYS_SYSPROTO_H_ 166struct sync_args { 167 int dummy; 168}; 169#endif 170/* ARGSUSED */ 171int 172sys_sync(struct thread *td, struct sync_args *uap) 173{ 174 175 return (kern_sync(td)); 176} 177 178/* 179 * Change filesystem quotas. 180 */ 181#ifndef _SYS_SYSPROTO_H_ 182struct quotactl_args { 183 char *path; 184 int cmd; 185 int uid; 186 caddr_t arg; 187}; 188#endif 189int 190sys_quotactl(struct thread *td, struct quotactl_args *uap) 191{ 192 struct mount *mp; 193 struct nameidata nd; 194 int error; 195 bool mp_busy; 196 197 AUDIT_ARG_CMD(uap->cmd); 198 AUDIT_ARG_UID(uap->uid); 199 if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS)) 200 return (EPERM); 201 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE, 202 uap->path); 203 if ((error = namei(&nd)) != 0) 204 return (error); 205 NDFREE_PNBUF(&nd); 206 mp = nd.ni_vp->v_mount; 207 vfs_ref(mp); 208 vput(nd.ni_vp); 209 error = vfs_busy(mp, 0); 210 if (error != 0) { 211 vfs_rel(mp); 212 return (error); 213 } 214 mp_busy = true; 215 error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg, &mp_busy); 216 217 /* 218 * Since quota on/off operations typically need to open quota 219 * files, the implementation may need to unbusy the mount point 220 * before calling into namei. Otherwise, unmount might be 221 * started between two vfs_busy() invocations (first is ours, 222 * second is from mount point cross-walk code in lookup()), 223 * causing deadlock. 224 * 225 * Avoid unbusying mp if the implementation indicates it has 226 * already done so. 227 */ 228 if (mp_busy) 229 vfs_unbusy(mp); 230 vfs_rel(mp); 231 return (error); 232} 233 234/* 235 * Used by statfs conversion routines to scale the block size up if 236 * necessary so that all of the block counts are <= 'max_size'. Note 237 * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero 238 * value of 'n'. 239 */ 240void 241statfs_scale_blocks(struct statfs *sf, long max_size) 242{ 243 uint64_t count; 244 int shift; 245 246 KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__)); 247 248 /* 249 * Attempt to scale the block counts to give a more accurate 250 * overview to userland of the ratio of free space to used 251 * space. To do this, find the largest block count and compute 252 * a divisor that lets it fit into a signed integer <= max_size. 253 */ 254 if (sf->f_bavail < 0) 255 count = -sf->f_bavail; 256 else 257 count = sf->f_bavail; 258 count = MAX(sf->f_blocks, MAX(sf->f_bfree, count)); 259 if (count <= max_size) 260 return; 261 262 count >>= flsl(max_size); 263 shift = 0; 264 while (count > 0) { 265 shift++; 266 count >>=1; 267 } 268 269 sf->f_bsize <<= shift; 270 sf->f_blocks >>= shift; 271 sf->f_bfree >>= shift; 272 sf->f_bavail >>= shift; 273} 274 275static int 276kern_do_statfs(struct thread *td, struct mount *mp, struct statfs *buf) 277{ 278 int error; 279 280 if (mp == NULL) 281 return (EBADF); 282 error = vfs_busy(mp, 0); 283 vfs_rel(mp); 284 if (error != 0) 285 return (error); 286#ifdef MAC 287 error = mac_mount_check_stat(td->td_ucred, mp); 288 if (error != 0) 289 goto out; 290#endif 291 error = VFS_STATFS(mp, buf); 292 if (error != 0) 293 goto out; 294 if (priv_check_cred_vfs_generation(td->td_ucred)) { 295 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0; 296 prison_enforce_statfs(td->td_ucred, mp, buf); 297 } 298out: 299 vfs_unbusy(mp); 300 return (error); 301} 302 303/* 304 * Get filesystem statistics. 305 */ 306#ifndef _SYS_SYSPROTO_H_ 307struct statfs_args { 308 char *path; 309 struct statfs *buf; 310}; 311#endif 312int 313sys_statfs(struct thread *td, struct statfs_args *uap) 314{ 315 struct statfs *sfp; 316 int error; 317 318 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 319 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp); 320 if (error == 0) 321 error = copyout(sfp, uap->buf, sizeof(struct statfs)); 322 free(sfp, M_STATFS); 323 return (error); 324} 325 326int 327kern_statfs(struct thread *td, const char *path, enum uio_seg pathseg, 328 struct statfs *buf) 329{ 330 struct mount *mp; 331 struct nameidata nd; 332 int error; 333 334 NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path); 335 error = namei(&nd); 336 if (error != 0) 337 return (error); 338 NDFREE_PNBUF(&nd); 339 mp = vfs_ref_from_vp(nd.ni_vp); 340 vrele(nd.ni_vp); 341 return (kern_do_statfs(td, mp, buf)); 342} 343 344/* 345 * Get filesystem statistics. 346 */ 347#ifndef _SYS_SYSPROTO_H_ 348struct fstatfs_args { 349 int fd; 350 struct statfs *buf; 351}; 352#endif 353int 354sys_fstatfs(struct thread *td, struct fstatfs_args *uap) 355{ 356 struct statfs *sfp; 357 int error; 358 359 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 360 error = kern_fstatfs(td, uap->fd, sfp); 361 if (error == 0) 362 error = copyout(sfp, uap->buf, sizeof(struct statfs)); 363 free(sfp, M_STATFS); 364 return (error); 365} 366 367int 368kern_fstatfs(struct thread *td, int fd, struct statfs *buf) 369{ 370 struct file *fp; 371 struct mount *mp; 372 struct vnode *vp; 373 int error; 374 375 AUDIT_ARG_FD(fd); 376 error = getvnode_path(td, fd, &cap_fstatfs_rights, &fp); 377 if (error != 0) 378 return (error); 379 vp = fp->f_vnode; 380#ifdef AUDIT 381 if (AUDITING_TD(td)) { 382 vn_lock(vp, LK_SHARED | LK_RETRY); 383 AUDIT_ARG_VNODE1(vp); 384 VOP_UNLOCK(vp); 385 } 386#endif 387 mp = vfs_ref_from_vp(vp); 388 fdrop(fp, td); 389 return (kern_do_statfs(td, mp, buf)); 390} 391 392/* 393 * Get statistics on all filesystems. 394 */ 395#ifndef _SYS_SYSPROTO_H_ 396struct getfsstat_args { 397 struct statfs *buf; 398 long bufsize; 399 int mode; 400}; 401#endif 402int 403sys_getfsstat(struct thread *td, struct getfsstat_args *uap) 404{ 405 size_t count; 406 int error; 407 408 if (uap->bufsize < 0 || uap->bufsize > SIZE_MAX) 409 return (EINVAL); 410 error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count, 411 UIO_USERSPACE, uap->mode); 412 if (error == 0) 413 td->td_retval[0] = count; 414 return (error); 415} 416 417/* 418 * If (bufsize > 0 && bufseg == UIO_SYSSPACE) 419 * The caller is responsible for freeing memory which will be allocated 420 * in '*buf'. 421 */ 422int 423kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize, 424 size_t *countp, enum uio_seg bufseg, int mode) 425{ 426 struct mount *mp, *nmp; 427 struct statfs *sfsp, *sp, *sptmp, *tofree; 428 size_t count, maxcount; 429 int error; 430 431 switch (mode) { 432 case MNT_WAIT: 433 case MNT_NOWAIT: 434 break; 435 default: 436 if (bufseg == UIO_SYSSPACE) 437 *buf = NULL; 438 return (EINVAL); 439 } 440restart: 441 maxcount = bufsize / sizeof(struct statfs); 442 if (bufsize == 0) { 443 sfsp = NULL; 444 tofree = NULL; 445 } else if (bufseg == UIO_USERSPACE) { 446 sfsp = *buf; 447 tofree = NULL; 448 } else /* if (bufseg == UIO_SYSSPACE) */ { 449 count = 0; 450 mtx_lock(&mountlist_mtx); 451 TAILQ_FOREACH(mp, &mountlist, mnt_list) { 452 count++; 453 } 454 mtx_unlock(&mountlist_mtx); 455 if (maxcount > count) 456 maxcount = count; 457 tofree = sfsp = *buf = malloc(maxcount * sizeof(struct statfs), 458 M_STATFS, M_WAITOK); 459 } 460 461 count = 0; 462 463 /* 464 * If there is no target buffer they only want the count. 465 * 466 * This could be TAILQ_FOREACH but it is open-coded to match the original 467 * code below. 468 */ 469 if (sfsp == NULL) { 470 mtx_lock(&mountlist_mtx); 471 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) { 472 if (prison_canseemount(td->td_ucred, mp) != 0) { 473 nmp = TAILQ_NEXT(mp, mnt_list); 474 continue; 475 } 476#ifdef MAC 477 if (mac_mount_check_stat(td->td_ucred, mp) != 0) { 478 nmp = TAILQ_NEXT(mp, mnt_list); 479 continue; 480 } 481#endif 482 count++; 483 nmp = TAILQ_NEXT(mp, mnt_list); 484 } 485 mtx_unlock(&mountlist_mtx); 486 *countp = count; 487 return (0); 488 } 489 490 /* 491 * They want the entire thing. 492 * 493 * Short-circuit the corner case of no room for anything, avoids 494 * relocking below. 495 */ 496 if (maxcount < 1) { 497 goto out; 498 } 499 500 mtx_lock(&mountlist_mtx); 501 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) { 502 if (prison_canseemount(td->td_ucred, mp) != 0) { 503 nmp = TAILQ_NEXT(mp, mnt_list); 504 continue; 505 } 506#ifdef MAC 507 if (mac_mount_check_stat(td->td_ucred, mp) != 0) { 508 nmp = TAILQ_NEXT(mp, mnt_list); 509 continue; 510 } 511#endif 512 if (mode == MNT_WAIT) { 513 if (vfs_busy(mp, MBF_MNTLSTLOCK) != 0) { 514 /* 515 * If vfs_busy() failed, and MBF_NOWAIT 516 * wasn't passed, then the mp is gone. 517 * Furthermore, because of MBF_MNTLSTLOCK, 518 * the mountlist_mtx was dropped. We have 519 * no other choice than to start over. 520 */ 521 mtx_unlock(&mountlist_mtx); 522 free(tofree, M_STATFS); 523 goto restart; 524 } 525 } else { 526 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK) != 0) { 527 nmp = TAILQ_NEXT(mp, mnt_list); 528 continue; 529 } 530 } 531 sp = &mp->mnt_stat; 532 /* 533 * If MNT_NOWAIT is specified, do not refresh 534 * the fsstat cache. 535 */ 536 if (mode != MNT_NOWAIT) { 537 error = VFS_STATFS(mp, sp); 538 if (error != 0) { 539 mtx_lock(&mountlist_mtx); 540 nmp = TAILQ_NEXT(mp, mnt_list); 541 vfs_unbusy(mp); 542 continue; 543 } 544 } 545 if (priv_check_cred_vfs_generation(td->td_ucred)) { 546 sptmp = malloc(sizeof(struct statfs), M_STATFS, 547 M_WAITOK); 548 *sptmp = *sp; 549 sptmp->f_fsid.val[0] = sptmp->f_fsid.val[1] = 0; 550 prison_enforce_statfs(td->td_ucred, mp, sptmp); 551 sp = sptmp; 552 } else 553 sptmp = NULL; 554 if (bufseg == UIO_SYSSPACE) { 555 bcopy(sp, sfsp, sizeof(*sp)); 556 free(sptmp, M_STATFS); 557 } else /* if (bufseg == UIO_USERSPACE) */ { 558 error = copyout(sp, sfsp, sizeof(*sp)); 559 free(sptmp, M_STATFS); 560 if (error != 0) { 561 vfs_unbusy(mp); 562 return (error); 563 } 564 } 565 sfsp++; 566 count++; 567 568 if (count == maxcount) { 569 vfs_unbusy(mp); 570 goto out; 571 } 572 573 mtx_lock(&mountlist_mtx); 574 nmp = TAILQ_NEXT(mp, mnt_list); 575 vfs_unbusy(mp); 576 } 577 mtx_unlock(&mountlist_mtx); 578out: 579 *countp = count; 580 return (0); 581} 582 583#ifdef COMPAT_FREEBSD4 584/* 585 * Get old format filesystem statistics. 586 */ 587static void freebsd4_cvtstatfs(struct statfs *, struct ostatfs *); 588 589#ifndef _SYS_SYSPROTO_H_ 590struct freebsd4_statfs_args { 591 char *path; 592 struct ostatfs *buf; 593}; 594#endif 595int 596freebsd4_statfs(struct thread *td, struct freebsd4_statfs_args *uap) 597{ 598 struct ostatfs osb; 599 struct statfs *sfp; 600 int error; 601 602 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 603 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp); 604 if (error == 0) { 605 freebsd4_cvtstatfs(sfp, &osb); 606 error = copyout(&osb, uap->buf, sizeof(osb)); 607 } 608 free(sfp, M_STATFS); 609 return (error); 610} 611 612/* 613 * Get filesystem statistics. 614 */ 615#ifndef _SYS_SYSPROTO_H_ 616struct freebsd4_fstatfs_args { 617 int fd; 618 struct ostatfs *buf; 619}; 620#endif 621int 622freebsd4_fstatfs(struct thread *td, struct freebsd4_fstatfs_args *uap) 623{ 624 struct ostatfs osb; 625 struct statfs *sfp; 626 int error; 627 628 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 629 error = kern_fstatfs(td, uap->fd, sfp); 630 if (error == 0) { 631 freebsd4_cvtstatfs(sfp, &osb); 632 error = copyout(&osb, uap->buf, sizeof(osb)); 633 } 634 free(sfp, M_STATFS); 635 return (error); 636} 637 638/* 639 * Get statistics on all filesystems. 640 */ 641#ifndef _SYS_SYSPROTO_H_ 642struct freebsd4_getfsstat_args { 643 struct ostatfs *buf; 644 long bufsize; 645 int mode; 646}; 647#endif 648int 649freebsd4_getfsstat(struct thread *td, struct freebsd4_getfsstat_args *uap) 650{ 651 struct statfs *buf, *sp; 652 struct ostatfs osb; 653 size_t count, size; 654 int error; 655 656 if (uap->bufsize < 0) 657 return (EINVAL); 658 count = uap->bufsize / sizeof(struct ostatfs); 659 if (count > SIZE_MAX / sizeof(struct statfs)) 660 return (EINVAL); 661 size = count * sizeof(struct statfs); 662 error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE, 663 uap->mode); 664 if (error == 0) 665 td->td_retval[0] = count; 666 if (size != 0) { 667 sp = buf; 668 while (count != 0 && error == 0) { 669 freebsd4_cvtstatfs(sp, &osb); 670 error = copyout(&osb, uap->buf, sizeof(osb)); 671 sp++; 672 uap->buf++; 673 count--; 674 } 675 free(buf, M_STATFS); 676 } 677 return (error); 678} 679 680/* 681 * Implement fstatfs() for (NFS) file handles. 682 */ 683#ifndef _SYS_SYSPROTO_H_ 684struct freebsd4_fhstatfs_args { 685 struct fhandle *u_fhp; 686 struct ostatfs *buf; 687}; 688#endif 689int 690freebsd4_fhstatfs(struct thread *td, struct freebsd4_fhstatfs_args *uap) 691{ 692 struct ostatfs osb; 693 struct statfs *sfp; 694 fhandle_t fh; 695 int error; 696 697 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t)); 698 if (error != 0) 699 return (error); 700 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 701 error = kern_fhstatfs(td, fh, sfp); 702 if (error == 0) { 703 freebsd4_cvtstatfs(sfp, &osb); 704 error = copyout(&osb, uap->buf, sizeof(osb)); 705 } 706 free(sfp, M_STATFS); 707 return (error); 708} 709 710/* 711 * Convert a new format statfs structure to an old format statfs structure. 712 */ 713static void 714freebsd4_cvtstatfs(struct statfs *nsp, struct ostatfs *osp) 715{ 716 717 statfs_scale_blocks(nsp, LONG_MAX); 718 bzero(osp, sizeof(*osp)); 719 osp->f_bsize = nsp->f_bsize; 720 osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX); 721 osp->f_blocks = nsp->f_blocks; 722 osp->f_bfree = nsp->f_bfree; 723 osp->f_bavail = nsp->f_bavail; 724 osp->f_files = MIN(nsp->f_files, LONG_MAX); 725 osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX); 726 osp->f_owner = nsp->f_owner; 727 osp->f_type = nsp->f_type; 728 osp->f_flags = nsp->f_flags; 729 osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX); 730 osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX); 731 osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX); 732 osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX); 733 strlcpy(osp->f_fstypename, nsp->f_fstypename, 734 MIN(MFSNAMELEN, OMFSNAMELEN)); 735 strlcpy(osp->f_mntonname, nsp->f_mntonname, 736 MIN(MNAMELEN, OMNAMELEN)); 737 strlcpy(osp->f_mntfromname, nsp->f_mntfromname, 738 MIN(MNAMELEN, OMNAMELEN)); 739 osp->f_fsid = nsp->f_fsid; 740} 741#endif /* COMPAT_FREEBSD4 */ 742 743#if defined(COMPAT_FREEBSD11) 744/* 745 * Get old format filesystem statistics. 746 */ 747static void freebsd11_cvtstatfs(struct statfs *, struct freebsd11_statfs *); 748 749int 750freebsd11_statfs(struct thread *td, struct freebsd11_statfs_args *uap) 751{ 752 struct freebsd11_statfs osb; 753 struct statfs *sfp; 754 int error; 755 756 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 757 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp); 758 if (error == 0) { 759 freebsd11_cvtstatfs(sfp, &osb); 760 error = copyout(&osb, uap->buf, sizeof(osb)); 761 } 762 free(sfp, M_STATFS); 763 return (error); 764} 765 766/* 767 * Get filesystem statistics. 768 */ 769int 770freebsd11_fstatfs(struct thread *td, struct freebsd11_fstatfs_args *uap) 771{ 772 struct freebsd11_statfs osb; 773 struct statfs *sfp; 774 int error; 775 776 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 777 error = kern_fstatfs(td, uap->fd, sfp); 778 if (error == 0) { 779 freebsd11_cvtstatfs(sfp, &osb); 780 error = copyout(&osb, uap->buf, sizeof(osb)); 781 } 782 free(sfp, M_STATFS); 783 return (error); 784} 785 786/* 787 * Get statistics on all filesystems. 788 */ 789int 790freebsd11_getfsstat(struct thread *td, struct freebsd11_getfsstat_args *uap) 791{ 792 return (kern_freebsd11_getfsstat(td, uap->buf, uap->bufsize, uap->mode)); 793} 794 795int 796kern_freebsd11_getfsstat(struct thread *td, struct freebsd11_statfs * ubuf, 797 long bufsize, int mode) 798{ 799 struct freebsd11_statfs osb; 800 struct statfs *buf, *sp; 801 size_t count, size; 802 int error; 803 804 if (bufsize < 0) 805 return (EINVAL); 806 807 count = bufsize / sizeof(struct ostatfs); 808 size = count * sizeof(struct statfs); 809 error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE, mode); 810 if (error == 0) 811 td->td_retval[0] = count; 812 if (size > 0) { 813 sp = buf; 814 while (count > 0 && error == 0) { 815 freebsd11_cvtstatfs(sp, &osb); 816 error = copyout(&osb, ubuf, sizeof(osb)); 817 sp++; 818 ubuf++; 819 count--; 820 } 821 free(buf, M_STATFS); 822 } 823 return (error); 824} 825 826/* 827 * Implement fstatfs() for (NFS) file handles. 828 */ 829int 830freebsd11_fhstatfs(struct thread *td, struct freebsd11_fhstatfs_args *uap) 831{ 832 struct freebsd11_statfs osb; 833 struct statfs *sfp; 834 fhandle_t fh; 835 int error; 836 837 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t)); 838 if (error) 839 return (error); 840 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 841 error = kern_fhstatfs(td, fh, sfp); 842 if (error == 0) { 843 freebsd11_cvtstatfs(sfp, &osb); 844 error = copyout(&osb, uap->buf, sizeof(osb)); 845 } 846 free(sfp, M_STATFS); 847 return (error); 848} 849 850/* 851 * Convert a new format statfs structure to an old format statfs structure. 852 */ 853static void 854freebsd11_cvtstatfs(struct statfs *nsp, struct freebsd11_statfs *osp) 855{ 856 857 bzero(osp, sizeof(*osp)); 858 osp->f_version = FREEBSD11_STATFS_VERSION; 859 osp->f_type = nsp->f_type; 860 osp->f_flags = nsp->f_flags; 861 osp->f_bsize = nsp->f_bsize; 862 osp->f_iosize = nsp->f_iosize; 863 osp->f_blocks = nsp->f_blocks; 864 osp->f_bfree = nsp->f_bfree; 865 osp->f_bavail = nsp->f_bavail; 866 osp->f_files = nsp->f_files; 867 osp->f_ffree = nsp->f_ffree; 868 osp->f_syncwrites = nsp->f_syncwrites; 869 osp->f_asyncwrites = nsp->f_asyncwrites; 870 osp->f_syncreads = nsp->f_syncreads; 871 osp->f_asyncreads = nsp->f_asyncreads; 872 osp->f_namemax = nsp->f_namemax; 873 osp->f_owner = nsp->f_owner; 874 osp->f_fsid = nsp->f_fsid; 875 strlcpy(osp->f_fstypename, nsp->f_fstypename, 876 MIN(MFSNAMELEN, sizeof(osp->f_fstypename))); 877 strlcpy(osp->f_mntonname, nsp->f_mntonname, 878 MIN(MNAMELEN, sizeof(osp->f_mntonname))); 879 strlcpy(osp->f_mntfromname, nsp->f_mntfromname, 880 MIN(MNAMELEN, sizeof(osp->f_mntfromname))); 881} 882#endif /* COMPAT_FREEBSD11 */ 883 884/* 885 * Change current working directory to a given file descriptor. 886 */ 887#ifndef _SYS_SYSPROTO_H_ 888struct fchdir_args { 889 int fd; 890}; 891#endif 892int 893sys_fchdir(struct thread *td, struct fchdir_args *uap) 894{ 895 struct vnode *vp, *tdp; 896 struct mount *mp; 897 struct file *fp; 898 int error; 899 900 AUDIT_ARG_FD(uap->fd); 901 error = getvnode_path(td, uap->fd, &cap_fchdir_rights, 902 &fp); 903 if (error != 0) 904 return (error); 905 vp = fp->f_vnode; 906 vrefact(vp); 907 fdrop(fp, td); 908 vn_lock(vp, LK_SHARED | LK_RETRY); 909 AUDIT_ARG_VNODE1(vp); 910 error = change_dir(vp, td); 911 while (!error && (mp = vp->v_mountedhere) != NULL) { 912 if (vfs_busy(mp, 0)) 913 continue; 914 error = VFS_ROOT(mp, LK_SHARED, &tdp); 915 vfs_unbusy(mp); 916 if (error != 0) 917 break; 918 vput(vp); 919 vp = tdp; 920 } 921 if (error != 0) { 922 vput(vp); 923 return (error); 924 } 925 VOP_UNLOCK(vp); 926 pwd_chdir(td, vp); 927 return (0); 928} 929 930/* 931 * Change current working directory (``.''). 932 */ 933#ifndef _SYS_SYSPROTO_H_ 934struct chdir_args { 935 char *path; 936}; 937#endif 938int 939sys_chdir(struct thread *td, struct chdir_args *uap) 940{ 941 942 return (kern_chdir(td, uap->path, UIO_USERSPACE)); 943} 944 945int 946kern_chdir(struct thread *td, const char *path, enum uio_seg pathseg) 947{ 948 struct nameidata nd; 949 int error; 950 951 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1, 952 pathseg, path); 953 if ((error = namei(&nd)) != 0) 954 return (error); 955 if ((error = change_dir(nd.ni_vp, td)) != 0) { 956 vput(nd.ni_vp); 957 NDFREE_PNBUF(&nd); 958 return (error); 959 } 960 VOP_UNLOCK(nd.ni_vp); 961 NDFREE_PNBUF(&nd); 962 pwd_chdir(td, nd.ni_vp); 963 return (0); 964} 965 966static int unprivileged_chroot = 0; 967SYSCTL_INT(_security_bsd, OID_AUTO, unprivileged_chroot, CTLFLAG_RW, 968 &unprivileged_chroot, 0, 969 "Unprivileged processes can use chroot(2)"); 970/* 971 * Change notion of root (``/'') directory. 972 */ 973#ifndef _SYS_SYSPROTO_H_ 974struct chroot_args { 975 char *path; 976}; 977#endif 978int 979sys_chroot(struct thread *td, struct chroot_args *uap) 980{ 981 struct nameidata nd; 982 struct proc *p; 983 int error; 984 985 error = priv_check(td, PRIV_VFS_CHROOT); 986 if (error != 0) { 987 p = td->td_proc; 988 PROC_LOCK(p); 989 if (unprivileged_chroot == 0 || 990 (p->p_flag2 & P2_NO_NEW_PRIVS) == 0) { 991 PROC_UNLOCK(p); 992 return (error); 993 } 994 PROC_UNLOCK(p); 995 } 996 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1, 997 UIO_USERSPACE, uap->path); 998 error = namei(&nd); 999 if (error != 0) 1000 return (error); 1001 NDFREE_PNBUF(&nd); 1002 error = change_dir(nd.ni_vp, td); 1003 if (error != 0) 1004 goto e_vunlock; 1005#ifdef MAC 1006 error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp); 1007 if (error != 0) 1008 goto e_vunlock; 1009#endif 1010 VOP_UNLOCK(nd.ni_vp); 1011 error = pwd_chroot(td, nd.ni_vp); 1012 vrele(nd.ni_vp); 1013 return (error); 1014e_vunlock: 1015 vput(nd.ni_vp); 1016 return (error); 1017} 1018 1019/* 1020 * Common routine for chroot and chdir. Callers must provide a locked vnode 1021 * instance. 1022 */ 1023int 1024change_dir(struct vnode *vp, struct thread *td) 1025{ 1026#ifdef MAC 1027 int error; 1028#endif 1029 1030 ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked"); 1031 if (vp->v_type != VDIR) 1032 return (ENOTDIR); 1033#ifdef MAC 1034 error = mac_vnode_check_chdir(td->td_ucred, vp); 1035 if (error != 0) 1036 return (error); 1037#endif 1038 return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td)); 1039} 1040 1041static __inline void 1042flags_to_rights(int flags, cap_rights_t *rightsp) 1043{ 1044 if (flags & O_EXEC) { 1045 cap_rights_set_one(rightsp, CAP_FEXECVE); 1046 if (flags & O_PATH) 1047 return; 1048 } else { 1049 switch ((flags & O_ACCMODE)) { 1050 case O_RDONLY: 1051 cap_rights_set_one(rightsp, CAP_READ); 1052 break; 1053 case O_RDWR: 1054 cap_rights_set_one(rightsp, CAP_READ); 1055 /* FALLTHROUGH */ 1056 case O_WRONLY: 1057 cap_rights_set_one(rightsp, CAP_WRITE); 1058 if (!(flags & (O_APPEND | O_TRUNC))) 1059 cap_rights_set_one(rightsp, CAP_SEEK); 1060 break; 1061 } 1062 } 1063 1064 if (flags & O_CREAT) 1065 cap_rights_set_one(rightsp, CAP_CREATE); 1066 1067 if (flags & O_TRUNC) 1068 cap_rights_set_one(rightsp, CAP_FTRUNCATE); 1069 1070 if (flags & (O_SYNC | O_FSYNC)) 1071 cap_rights_set_one(rightsp, CAP_FSYNC); 1072 1073 if (flags & (O_EXLOCK | O_SHLOCK)) 1074 cap_rights_set_one(rightsp, CAP_FLOCK); 1075} 1076 1077/* 1078 * Check permissions, allocate an open file structure, and call the device 1079 * open routine if any. 1080 */ 1081#ifndef _SYS_SYSPROTO_H_ 1082struct open_args { 1083 char *path; 1084 int flags; 1085 int mode; 1086}; 1087#endif 1088int 1089sys_open(struct thread *td, struct open_args *uap) 1090{ 1091 1092 return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 1093 uap->flags, uap->mode)); 1094} 1095 1096#ifndef _SYS_SYSPROTO_H_ 1097struct openat_args { 1098 int fd; 1099 char *path; 1100 int flag; 1101 int mode; 1102}; 1103#endif 1104int 1105sys_openat(struct thread *td, struct openat_args *uap) 1106{ 1107 1108 AUDIT_ARG_FD(uap->fd); 1109 return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag, 1110 uap->mode)); 1111} 1112 1113/* 1114 * If fpp != NULL, opened file is not installed into the file 1115 * descriptor table, instead it is returned in *fpp. This is 1116 * incompatible with fdopen(), in which case we return EINVAL. 1117 */ 1118static int 1119openatfp(struct thread *td, int dirfd, const char *path, 1120 enum uio_seg pathseg, int flags, int mode, struct file **fpp) 1121{ 1122 struct proc *p; 1123 struct filedesc *fdp; 1124 struct pwddesc *pdp; 1125 struct file *fp; 1126 struct vnode *vp; 1127 struct filecaps *fcaps; 1128 struct nameidata nd; 1129 cap_rights_t rights; 1130 int cmode, error, indx; 1131 1132 indx = -1; 1133 p = td->td_proc; 1134 fdp = p->p_fd; 1135 pdp = p->p_pd; 1136 1137 AUDIT_ARG_FFLAGS(flags); 1138 AUDIT_ARG_MODE(mode); 1139 cap_rights_init_one(&rights, CAP_LOOKUP); 1140 flags_to_rights(flags, &rights); 1141 1142 /* 1143 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags 1144 * may be specified. On the other hand, for O_PATH any mode 1145 * except O_EXEC is ignored. 1146 */ 1147 if ((flags & O_PATH) != 0) { 1148 flags &= ~(O_CREAT | O_ACCMODE); 1149 } else if ((flags & O_EXEC) != 0) { 1150 if (flags & O_ACCMODE) 1151 return (EINVAL); 1152 } else if ((flags & O_ACCMODE) == O_ACCMODE) { 1153 return (EINVAL); 1154 } else { 1155 flags = FFLAGS(flags); 1156 } 1157 1158 /* 1159 * Allocate a file structure. The descriptor to reference it 1160 * is allocated and used by finstall_refed() below. 1161 */ 1162 error = falloc_noinstall(td, &fp); 1163 if (error != 0) 1164 return (error); 1165 /* Set the flags early so the finit in devfs can pick them up. */ 1166 fp->f_flag = flags & FMASK; 1167 cmode = ((mode & ~pdp->pd_cmask) & ALLPERMS) & ~S_ISTXT; 1168 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1 | WANTIOCTLCAPS, 1169 pathseg, path, dirfd, &rights); 1170 td->td_dupfd = -1; /* XXX check for fdopen */ 1171 error = vn_open_cred(&nd, &flags, cmode, VN_OPEN_WANTIOCTLCAPS, 1172 td->td_ucred, fp); 1173 if (error != 0) { 1174 /* 1175 * If the vn_open replaced the method vector, something 1176 * wonderous happened deep below and we just pass it up 1177 * pretending we know what we do. 1178 */ 1179 if (error == ENXIO && fp->f_ops != &badfileops) { 1180 MPASS((flags & O_PATH) == 0); 1181 goto success; 1182 } 1183 1184 /* 1185 * Handle special fdopen() case. bleh. 1186 * 1187 * Don't do this for relative (capability) lookups; we don't 1188 * understand exactly what would happen, and we don't think 1189 * that it ever should. 1190 */ 1191 if ((nd.ni_resflags & NIRES_STRICTREL) == 0 && 1192 (error == ENODEV || error == ENXIO) && 1193 td->td_dupfd >= 0) { 1194 MPASS(fpp == NULL); 1195 error = dupfdopen(td, fdp, td->td_dupfd, flags, error, 1196 &indx); 1197 if (error == 0) 1198 goto success; 1199 } 1200 1201 goto bad; 1202 } 1203 td->td_dupfd = 0; 1204 NDFREE_PNBUF(&nd); 1205 vp = nd.ni_vp; 1206 1207 /* 1208 * Store the vnode, for any f_type. Typically, the vnode use 1209 * count is decremented by direct call to vn_closefile() for 1210 * files that switched type in the cdevsw fdopen() method. 1211 */ 1212 fp->f_vnode = vp; 1213 1214 /* 1215 * If the file wasn't claimed by devfs bind it to the normal 1216 * vnode operations here. 1217 */ 1218 if (fp->f_ops == &badfileops) { 1219 KASSERT(vp->v_type != VFIFO || (flags & O_PATH) != 0, 1220 ("Unexpected fifo fp %p vp %p", fp, vp)); 1221 if ((flags & O_PATH) != 0) { 1222 finit(fp, (flags & FMASK) | (fp->f_flag & FKQALLOWED), 1223 DTYPE_VNODE, NULL, &path_fileops); 1224 } else { 1225 finit_vnode(fp, flags, NULL, &vnops); 1226 } 1227 } 1228 1229 VOP_UNLOCK(vp); 1230 if (flags & O_TRUNC) { 1231 error = fo_truncate(fp, 0, td->td_ucred, td); 1232 if (error != 0) 1233 goto bad; 1234 } 1235success: 1236 if (fpp != NULL) { 1237 MPASS(error == 0); 1238 NDFREE_IOCTLCAPS(&nd); 1239 *fpp = fp; 1240 return (0); 1241 } 1242 1243 /* 1244 * If we haven't already installed the FD (for dupfdopen), do so now. 1245 */ 1246 if (indx == -1) { 1247#ifdef CAPABILITIES 1248 if ((nd.ni_resflags & NIRES_STRICTREL) != 0) 1249 fcaps = &nd.ni_filecaps; 1250 else 1251#endif 1252 fcaps = NULL; 1253 error = finstall_refed(td, fp, &indx, flags, fcaps); 1254 /* On success finstall_refed() consumes fcaps. */ 1255 if (error != 0) { 1256 goto bad; 1257 } 1258 } else { 1259 NDFREE_IOCTLCAPS(&nd); 1260 falloc_abort(td, fp); 1261 } 1262 1263 td->td_retval[0] = indx; 1264 return (0); 1265bad: 1266 KASSERT(indx == -1, ("indx=%d, should be -1", indx)); 1267 NDFREE_IOCTLCAPS(&nd); 1268 falloc_abort(td, fp); 1269 return (error); 1270} 1271 1272int 1273kern_openat(struct thread *td, int dirfd, const char *path, 1274 enum uio_seg pathseg, int flags, int mode) 1275{ 1276 return (openatfp(td, dirfd, path, pathseg, flags, mode, NULL)); 1277} 1278 1279int 1280kern_openatfp(struct thread *td, int dirfd, const char *path, 1281 enum uio_seg pathseg, int flags, int mode, struct file **fpp) 1282{ 1283 int error, old_dupfd; 1284 1285 old_dupfd = td->td_dupfd; 1286 td->td_dupfd = -1; 1287 error = openatfp(td, dirfd, path, pathseg, flags, mode, fpp); 1288 td->td_dupfd = old_dupfd; 1289 return (error); 1290} 1291 1292#ifdef COMPAT_43 1293/* 1294 * Create a file. 1295 */ 1296#ifndef _SYS_SYSPROTO_H_ 1297struct ocreat_args { 1298 char *path; 1299 int mode; 1300}; 1301#endif 1302int 1303ocreat(struct thread *td, struct ocreat_args *uap) 1304{ 1305 1306 return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 1307 O_WRONLY | O_CREAT | O_TRUNC, uap->mode)); 1308} 1309#endif /* COMPAT_43 */ 1310 1311/* 1312 * Create a special file. 1313 */ 1314#ifndef _SYS_SYSPROTO_H_ 1315struct mknodat_args { 1316 int fd; 1317 char *path; 1318 mode_t mode; 1319 dev_t dev; 1320}; 1321#endif 1322int 1323sys_mknodat(struct thread *td, struct mknodat_args *uap) 1324{ 1325 1326 return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode, 1327 uap->dev)); 1328} 1329 1330#if defined(COMPAT_FREEBSD11) 1331int 1332freebsd11_mknod(struct thread *td, 1333 struct freebsd11_mknod_args *uap) 1334{ 1335 1336 return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 1337 uap->mode, uap->dev)); 1338} 1339 1340int 1341freebsd11_mknodat(struct thread *td, 1342 struct freebsd11_mknodat_args *uap) 1343{ 1344 1345 return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode, 1346 uap->dev)); 1347} 1348#endif /* COMPAT_FREEBSD11 */ 1349 1350int 1351kern_mknodat(struct thread *td, int fd, const char *path, enum uio_seg pathseg, 1352 int mode, dev_t dev) 1353{ 1354 struct vnode *vp; 1355 struct mount *mp; 1356 struct vattr vattr; 1357 struct nameidata nd; 1358 int error, whiteout = 0; 1359 1360 AUDIT_ARG_MODE(mode); 1361 AUDIT_ARG_DEV(dev); 1362 switch (mode & S_IFMT) { 1363 case S_IFCHR: 1364 case S_IFBLK: 1365 error = priv_check(td, PRIV_VFS_MKNOD_DEV); 1366 if (error == 0 && dev == VNOVAL) 1367 error = EINVAL; 1368 break; 1369 case S_IFWHT: 1370 error = priv_check(td, PRIV_VFS_MKNOD_WHT); 1371 break; 1372 case S_IFIFO: 1373 if (dev == 0) 1374 return (kern_mkfifoat(td, fd, path, pathseg, mode)); 1375 /* FALLTHROUGH */ 1376 default: 1377 error = EINVAL; 1378 break; 1379 } 1380 if (error != 0) 1381 return (error); 1382 NDPREINIT(&nd); 1383restart: 1384 bwillwrite(); 1385 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | AUDITVNODE1 | NOCACHE, 1386 pathseg, path, fd, &cap_mknodat_rights); 1387 if ((error = namei(&nd)) != 0) 1388 return (error); 1389 vp = nd.ni_vp; 1390 if (vp != NULL) { 1391 NDFREE_PNBUF(&nd); 1392 if (vp == nd.ni_dvp) 1393 vrele(nd.ni_dvp); 1394 else 1395 vput(nd.ni_dvp); 1396 vrele(vp); 1397 return (EEXIST); 1398 } else { 1399 VATTR_NULL(&vattr); 1400 vattr.va_mode = (mode & ALLPERMS) & 1401 ~td->td_proc->p_pd->pd_cmask; 1402 vattr.va_rdev = dev; 1403 whiteout = 0; 1404 1405 switch (mode & S_IFMT) { 1406 case S_IFCHR: 1407 vattr.va_type = VCHR; 1408 break; 1409 case S_IFBLK: 1410 vattr.va_type = VBLK; 1411 break; 1412 case S_IFWHT: 1413 whiteout = 1; 1414 break; 1415 default: 1416 panic("kern_mknod: invalid mode"); 1417 } 1418 } 1419 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1420 NDFREE_PNBUF(&nd); 1421 vput(nd.ni_dvp); 1422 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH)) != 0) 1423 return (error); 1424 goto restart; 1425 } 1426#ifdef MAC 1427 if (error == 0 && !whiteout) 1428 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, 1429 &nd.ni_cnd, &vattr); 1430#endif 1431 if (error == 0) { 1432 if (whiteout) 1433 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1434 else { 1435 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1436 &nd.ni_cnd, &vattr); 1437 } 1438 } 1439 VOP_VPUT_PAIR(nd.ni_dvp, error == 0 && !whiteout ? &nd.ni_vp : NULL, 1440 true); 1441 vn_finished_write(mp); 1442 NDFREE_PNBUF(&nd); 1443 if (error == ERELOOKUP) 1444 goto restart; 1445 return (error); 1446} 1447 1448/* 1449 * Create a named pipe. 1450 */ 1451#ifndef _SYS_SYSPROTO_H_ 1452struct mkfifo_args { 1453 char *path; 1454 int mode; 1455}; 1456#endif 1457int 1458sys_mkfifo(struct thread *td, struct mkfifo_args *uap) 1459{ 1460 1461 return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 1462 uap->mode)); 1463} 1464 1465#ifndef _SYS_SYSPROTO_H_ 1466struct mkfifoat_args { 1467 int fd; 1468 char *path; 1469 mode_t mode; 1470}; 1471#endif 1472int 1473sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap) 1474{ 1475 1476 return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE, 1477 uap->mode)); 1478} 1479 1480int 1481kern_mkfifoat(struct thread *td, int fd, const char *path, 1482 enum uio_seg pathseg, int mode) 1483{ 1484 struct mount *mp; 1485 struct vattr vattr; 1486 struct nameidata nd; 1487 int error; 1488 1489 AUDIT_ARG_MODE(mode); 1490 NDPREINIT(&nd); 1491restart: 1492 bwillwrite(); 1493 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | AUDITVNODE1 | NOCACHE, 1494 pathseg, path, fd, &cap_mkfifoat_rights); 1495 if ((error = namei(&nd)) != 0) 1496 return (error); 1497 if (nd.ni_vp != NULL) { 1498 NDFREE_PNBUF(&nd); 1499 if (nd.ni_vp == nd.ni_dvp) 1500 vrele(nd.ni_dvp); 1501 else 1502 vput(nd.ni_dvp); 1503 vrele(nd.ni_vp); 1504 return (EEXIST); 1505 } 1506 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1507 NDFREE_PNBUF(&nd); 1508 vput(nd.ni_dvp); 1509 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH)) != 0) 1510 return (error); 1511 goto restart; 1512 } 1513 VATTR_NULL(&vattr); 1514 vattr.va_type = VFIFO; 1515 vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_pd->pd_cmask; 1516#ifdef MAC 1517 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd, 1518 &vattr); 1519 if (error != 0) 1520 goto out; 1521#endif 1522 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1523#ifdef MAC 1524out: 1525#endif 1526 VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true); 1527 vn_finished_write(mp); 1528 NDFREE_PNBUF(&nd); 1529 if (error == ERELOOKUP) 1530 goto restart; 1531 return (error); 1532} 1533 1534/* 1535 * Make a hard file link. 1536 */ 1537#ifndef _SYS_SYSPROTO_H_ 1538struct link_args { 1539 char *path; 1540 char *link; 1541}; 1542#endif 1543int 1544sys_link(struct thread *td, struct link_args *uap) 1545{ 1546 1547 return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link, 1548 UIO_USERSPACE, AT_SYMLINK_FOLLOW)); 1549} 1550 1551#ifndef _SYS_SYSPROTO_H_ 1552struct linkat_args { 1553 int fd1; 1554 char *path1; 1555 int fd2; 1556 char *path2; 1557 int flag; 1558}; 1559#endif 1560int 1561sys_linkat(struct thread *td, struct linkat_args *uap) 1562{ 1563 1564 return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2, 1565 UIO_USERSPACE, uap->flag)); 1566} 1567 1568int hardlink_check_uid = 0; 1569SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW, 1570 &hardlink_check_uid, 0, 1571 "Unprivileged processes cannot create hard links to files owned by other " 1572 "users"); 1573static int hardlink_check_gid = 0; 1574SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW, 1575 &hardlink_check_gid, 0, 1576 "Unprivileged processes cannot create hard links to files owned by other " 1577 "groups"); 1578 1579static int 1580can_hardlink(struct vnode *vp, struct ucred *cred) 1581{ 1582 struct vattr va; 1583 int error; 1584 1585 if (!hardlink_check_uid && !hardlink_check_gid) 1586 return (0); 1587 1588 error = VOP_GETATTR(vp, &va, cred); 1589 if (error != 0) 1590 return (error); 1591 1592 if (hardlink_check_uid && cred->cr_uid != va.va_uid) { 1593 error = priv_check_cred(cred, PRIV_VFS_LINK); 1594 if (error != 0) 1595 return (error); 1596 } 1597 1598 if (hardlink_check_gid && !groupmember(va.va_gid, cred)) { 1599 error = priv_check_cred(cred, PRIV_VFS_LINK); 1600 if (error != 0) 1601 return (error); 1602 } 1603 1604 return (0); 1605} 1606 1607int 1608kern_linkat(struct thread *td, int fd1, int fd2, const char *path1, 1609 const char *path2, enum uio_seg segflag, int flag) 1610{ 1611 struct nameidata nd; 1612 int error; 1613 1614 if ((flag & ~(AT_SYMLINK_FOLLOW | AT_RESOLVE_BENEATH | 1615 AT_EMPTY_PATH)) != 0) 1616 return (EINVAL); 1617 1618 NDPREINIT(&nd); 1619 do { 1620 bwillwrite(); 1621 NDINIT_ATRIGHTS(&nd, LOOKUP, AUDITVNODE1 | at2cnpflags(flag, 1622 AT_SYMLINK_FOLLOW | AT_RESOLVE_BENEATH | AT_EMPTY_PATH), 1623 segflag, path1, fd1, &cap_linkat_source_rights); 1624 if ((error = namei(&nd)) != 0) 1625 return (error); 1626 NDFREE_PNBUF(&nd); 1627 if ((nd.ni_resflags & NIRES_EMPTYPATH) != 0) { 1628 error = priv_check(td, PRIV_VFS_FHOPEN); 1629 if (error != 0) { 1630 vrele(nd.ni_vp); 1631 return (error); 1632 } 1633 } 1634 error = kern_linkat_vp(td, nd.ni_vp, fd2, path2, segflag); 1635 } while (error == EAGAIN || error == ERELOOKUP); 1636 return (error); 1637} 1638 1639static int 1640kern_linkat_vp(struct thread *td, struct vnode *vp, int fd, const char *path, 1641 enum uio_seg segflag) 1642{ 1643 struct nameidata nd; 1644 struct mount *mp; 1645 int error; 1646 1647 if (vp->v_type == VDIR) { 1648 vrele(vp); 1649 return (EPERM); /* POSIX */ 1650 } 1651 NDINIT_ATRIGHTS(&nd, CREATE, 1652 LOCKPARENT | AUDITVNODE2 | NOCACHE, segflag, path, fd, 1653 &cap_linkat_target_rights); 1654 if ((error = namei(&nd)) == 0) { 1655 if (nd.ni_vp != NULL) { 1656 NDFREE_PNBUF(&nd); 1657 if (nd.ni_dvp == nd.ni_vp) 1658 vrele(nd.ni_dvp); 1659 else 1660 vput(nd.ni_dvp); 1661 vrele(nd.ni_vp); 1662 vrele(vp); 1663 return (EEXIST); 1664 } else if (nd.ni_dvp->v_mount != vp->v_mount) { 1665 /* 1666 * Cross-device link. No need to recheck 1667 * vp->v_type, since it cannot change, except 1668 * to VBAD. 1669 */ 1670 NDFREE_PNBUF(&nd); 1671 vput(nd.ni_dvp); 1672 vrele(vp); 1673 return (EXDEV); 1674 } else if (vn_lock(vp, LK_EXCLUSIVE) == 0) { 1675 error = can_hardlink(vp, td->td_ucred); 1676#ifdef MAC 1677 if (error == 0) 1678 error = mac_vnode_check_link(td->td_ucred, 1679 nd.ni_dvp, vp, &nd.ni_cnd); 1680#endif 1681 if (error != 0) { 1682 vput(vp); 1683 vput(nd.ni_dvp); 1684 NDFREE_PNBUF(&nd); 1685 return (error); 1686 } 1687 error = vn_start_write(vp, &mp, V_NOWAIT); 1688 if (error != 0) { 1689 vput(vp); 1690 vput(nd.ni_dvp); 1691 NDFREE_PNBUF(&nd); 1692 error = vn_start_write(NULL, &mp, 1693 V_XSLEEP | V_PCATCH); 1694 if (error != 0) 1695 return (error); 1696 return (EAGAIN); 1697 } 1698 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1699 VOP_VPUT_PAIR(nd.ni_dvp, &vp, true); 1700 vn_finished_write(mp); 1701 NDFREE_PNBUF(&nd); 1702 vp = NULL; 1703 } else { 1704 vput(nd.ni_dvp); 1705 NDFREE_PNBUF(&nd); 1706 vrele(vp); 1707 return (EAGAIN); 1708 } 1709 } 1710 if (vp != NULL) 1711 vrele(vp); 1712 return (error); 1713} 1714 1715/* 1716 * Make a symbolic link. 1717 */ 1718#ifndef _SYS_SYSPROTO_H_ 1719struct symlink_args { 1720 char *path; 1721 char *link; 1722}; 1723#endif 1724int 1725sys_symlink(struct thread *td, struct symlink_args *uap) 1726{ 1727 1728 return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link, 1729 UIO_USERSPACE)); 1730} 1731 1732#ifndef _SYS_SYSPROTO_H_ 1733struct symlinkat_args { 1734 char *path; 1735 int fd; 1736 char *path2; 1737}; 1738#endif 1739int 1740sys_symlinkat(struct thread *td, struct symlinkat_args *uap) 1741{ 1742 1743 return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2, 1744 UIO_USERSPACE)); 1745} 1746 1747int 1748kern_symlinkat(struct thread *td, const char *path1, int fd, const char *path2, 1749 enum uio_seg segflg) 1750{ 1751 struct mount *mp; 1752 struct vattr vattr; 1753 const char *syspath; 1754 char *tmppath; 1755 struct nameidata nd; 1756 int error; 1757 1758 if (segflg == UIO_SYSSPACE) { 1759 syspath = path1; 1760 } else { 1761 tmppath = uma_zalloc(namei_zone, M_WAITOK); 1762 if ((error = copyinstr(path1, tmppath, MAXPATHLEN, NULL)) != 0) 1763 goto out; 1764 syspath = tmppath; 1765 } 1766 AUDIT_ARG_TEXT(syspath); 1767 NDPREINIT(&nd); 1768restart: 1769 bwillwrite(); 1770 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | AUDITVNODE1 | NOCACHE, segflg, 1771 path2, fd, &cap_symlinkat_rights); 1772 if ((error = namei(&nd)) != 0) 1773 goto out; 1774 if (nd.ni_vp) { 1775 NDFREE_PNBUF(&nd); 1776 if (nd.ni_vp == nd.ni_dvp) 1777 vrele(nd.ni_dvp); 1778 else 1779 vput(nd.ni_dvp); 1780 vrele(nd.ni_vp); 1781 nd.ni_vp = NULL; 1782 error = EEXIST; 1783 goto out; 1784 } 1785 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1786 NDFREE_PNBUF(&nd); 1787 vput(nd.ni_dvp); 1788 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH)) != 0) 1789 goto out; 1790 goto restart; 1791 } 1792 VATTR_NULL(&vattr); 1793 vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_pd->pd_cmask; 1794#ifdef MAC 1795 vattr.va_type = VLNK; 1796 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd, 1797 &vattr); 1798 if (error != 0) 1799 goto out2; 1800#endif 1801 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath); 1802#ifdef MAC 1803out2: 1804#endif 1805 VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true); 1806 vn_finished_write(mp); 1807 NDFREE_PNBUF(&nd); 1808 if (error == ERELOOKUP) 1809 goto restart; 1810out: 1811 if (segflg != UIO_SYSSPACE) 1812 uma_zfree(namei_zone, tmppath); 1813 return (error); 1814} 1815 1816/* 1817 * Delete a whiteout from the filesystem. 1818 */ 1819#ifndef _SYS_SYSPROTO_H_ 1820struct undelete_args { 1821 char *path; 1822}; 1823#endif 1824int 1825sys_undelete(struct thread *td, struct undelete_args *uap) 1826{ 1827 struct mount *mp; 1828 struct nameidata nd; 1829 int error; 1830 1831 NDPREINIT(&nd); 1832restart: 1833 bwillwrite(); 1834 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1, 1835 UIO_USERSPACE, uap->path); 1836 error = namei(&nd); 1837 if (error != 0) 1838 return (error); 1839 1840 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 1841 NDFREE_PNBUF(&nd); 1842 if (nd.ni_vp == nd.ni_dvp) 1843 vrele(nd.ni_dvp); 1844 else 1845 vput(nd.ni_dvp); 1846 if (nd.ni_vp) 1847 vrele(nd.ni_vp); 1848 return (EEXIST); 1849 } 1850 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1851 NDFREE_PNBUF(&nd); 1852 vput(nd.ni_dvp); 1853 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH)) != 0) 1854 return (error); 1855 goto restart; 1856 } 1857 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE); 1858 NDFREE_PNBUF(&nd); 1859 vput(nd.ni_dvp); 1860 vn_finished_write(mp); 1861 if (error == ERELOOKUP) 1862 goto restart; 1863 return (error); 1864} 1865 1866/* 1867 * Delete a name from the filesystem. 1868 */ 1869#ifndef _SYS_SYSPROTO_H_ 1870struct unlink_args { 1871 char *path; 1872}; 1873#endif 1874int 1875sys_unlink(struct thread *td, struct unlink_args *uap) 1876{ 1877 1878 return (kern_funlinkat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE, 1879 0, 0)); 1880} 1881 1882static int 1883kern_funlinkat_ex(struct thread *td, int dfd, const char *path, int fd, 1884 int flag, enum uio_seg pathseg, ino_t oldinum) 1885{ 1886 1887 if ((flag & ~(AT_REMOVEDIR | AT_RESOLVE_BENEATH)) != 0) 1888 return (EINVAL); 1889 1890 if ((flag & AT_REMOVEDIR) != 0) 1891 return (kern_frmdirat(td, dfd, path, fd, UIO_USERSPACE, 0)); 1892 1893 return (kern_funlinkat(td, dfd, path, fd, UIO_USERSPACE, 0, 0)); 1894} 1895 1896#ifndef _SYS_SYSPROTO_H_ 1897struct unlinkat_args { 1898 int fd; 1899 char *path; 1900 int flag; 1901}; 1902#endif 1903int 1904sys_unlinkat(struct thread *td, struct unlinkat_args *uap) 1905{ 1906 1907 return (kern_funlinkat_ex(td, uap->fd, uap->path, FD_NONE, uap->flag, 1908 UIO_USERSPACE, 0)); 1909} 1910 1911#ifndef _SYS_SYSPROTO_H_ 1912struct funlinkat_args { 1913 int dfd; 1914 const char *path; 1915 int fd; 1916 int flag; 1917}; 1918#endif 1919int 1920sys_funlinkat(struct thread *td, struct funlinkat_args *uap) 1921{ 1922 1923 return (kern_funlinkat_ex(td, uap->dfd, uap->path, uap->fd, uap->flag, 1924 UIO_USERSPACE, 0)); 1925} 1926 1927int 1928kern_funlinkat(struct thread *td, int dfd, const char *path, int fd, 1929 enum uio_seg pathseg, int flag, ino_t oldinum) 1930{ 1931 struct mount *mp; 1932 struct file *fp; 1933 struct vnode *vp; 1934 struct nameidata nd; 1935 struct stat sb; 1936 int error; 1937 1938 fp = NULL; 1939 if (fd != FD_NONE) { 1940 error = getvnode_path(td, fd, &cap_no_rights, &fp); 1941 if (error != 0) 1942 return (error); 1943 } 1944 1945 NDPREINIT(&nd); 1946restart: 1947 bwillwrite(); 1948 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 | 1949 at2cnpflags(flag, AT_RESOLVE_BENEATH), 1950 pathseg, path, dfd, &cap_unlinkat_rights); 1951 if ((error = namei(&nd)) != 0) { 1952 if (error == EINVAL) 1953 error = EPERM; 1954 goto fdout; 1955 } 1956 vp = nd.ni_vp; 1957 if (vp->v_type == VDIR && oldinum == 0) { 1958 error = EPERM; /* POSIX */ 1959 } else if (oldinum != 0 && 1960 ((error = VOP_STAT(vp, &sb, td->td_ucred, NOCRED)) == 0) && 1961 sb.st_ino != oldinum) { 1962 error = EIDRM; /* Identifier removed */ 1963 } else if (fp != NULL && fp->f_vnode != vp) { 1964 if (VN_IS_DOOMED(fp->f_vnode)) 1965 error = EBADF; 1966 else 1967 error = EDEADLK; 1968 } else { 1969 /* 1970 * The root of a mounted filesystem cannot be deleted. 1971 * 1972 * XXX: can this only be a VDIR case? 1973 */ 1974 if (vp->v_vflag & VV_ROOT) 1975 error = EBUSY; 1976 } 1977 if (error == 0) { 1978 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1979 NDFREE_PNBUF(&nd); 1980 vput(nd.ni_dvp); 1981 if (vp == nd.ni_dvp) 1982 vrele(vp); 1983 else 1984 vput(vp); 1985 if ((error = vn_start_write(NULL, &mp, 1986 V_XSLEEP | V_PCATCH)) != 0) { 1987 goto fdout; 1988 } 1989 goto restart; 1990 } 1991#ifdef MAC 1992 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp, 1993 &nd.ni_cnd); 1994 if (error != 0) 1995 goto out; 1996#endif 1997 vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK); 1998 error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd); 1999#ifdef MAC 2000out: 2001#endif 2002 vn_finished_write(mp); 2003 } 2004 NDFREE_PNBUF(&nd); 2005 vput(nd.ni_dvp); 2006 if (vp == nd.ni_dvp) 2007 vrele(vp); 2008 else 2009 vput(vp); 2010 if (error == ERELOOKUP) 2011 goto restart; 2012fdout: 2013 if (fp != NULL) 2014 fdrop(fp, td); 2015 return (error); 2016} 2017 2018/* 2019 * Reposition read/write file offset. 2020 */ 2021#ifndef _SYS_SYSPROTO_H_ 2022struct lseek_args { 2023 int fd; 2024 int pad; 2025 off_t offset; 2026 int whence; 2027}; 2028#endif 2029int 2030sys_lseek(struct thread *td, struct lseek_args *uap) 2031{ 2032 2033 return (kern_lseek(td, uap->fd, uap->offset, uap->whence)); 2034} 2035 2036int 2037kern_lseek(struct thread *td, int fd, off_t offset, int whence) 2038{ 2039 struct file *fp; 2040 int error; 2041 2042 AUDIT_ARG_FD(fd); 2043 error = fget(td, fd, &cap_seek_rights, &fp); 2044 if (error != 0) 2045 return (error); 2046 error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ? 2047 fo_seek(fp, offset, whence, td) : ESPIPE; 2048 fdrop(fp, td); 2049 return (error); 2050} 2051 2052#if defined(COMPAT_43) 2053/* 2054 * Reposition read/write file offset. 2055 */ 2056#ifndef _SYS_SYSPROTO_H_ 2057struct olseek_args { 2058 int fd; 2059 long offset; 2060 int whence; 2061}; 2062#endif 2063int 2064olseek(struct thread *td, struct olseek_args *uap) 2065{ 2066 2067 return (kern_lseek(td, uap->fd, uap->offset, uap->whence)); 2068} 2069#endif /* COMPAT_43 */ 2070 2071#if defined(COMPAT_FREEBSD6) 2072/* Version with the 'pad' argument */ 2073int 2074freebsd6_lseek(struct thread *td, struct freebsd6_lseek_args *uap) 2075{ 2076 2077 return (kern_lseek(td, uap->fd, uap->offset, uap->whence)); 2078} 2079#endif 2080 2081/* 2082 * Check access permissions using passed credentials. 2083 */ 2084static int 2085vn_access(struct vnode *vp, int user_flags, struct ucred *cred, 2086 struct thread *td) 2087{ 2088 accmode_t accmode; 2089 int error; 2090 2091 /* Flags == 0 means only check for existence. */ 2092 if (user_flags == 0) 2093 return (0); 2094 2095 accmode = 0; 2096 if (user_flags & R_OK) 2097 accmode |= VREAD; 2098 if (user_flags & W_OK) 2099 accmode |= VWRITE; 2100 if (user_flags & X_OK) 2101 accmode |= VEXEC; 2102#ifdef MAC 2103 error = mac_vnode_check_access(cred, vp, accmode); 2104 if (error != 0) 2105 return (error); 2106#endif 2107 if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0) 2108 error = VOP_ACCESS(vp, accmode, cred, td); 2109 return (error); 2110} 2111 2112/* 2113 * Check access permissions using "real" credentials. 2114 */ 2115#ifndef _SYS_SYSPROTO_H_ 2116struct access_args { 2117 char *path; 2118 int amode; 2119}; 2120#endif 2121int 2122sys_access(struct thread *td, struct access_args *uap) 2123{ 2124 2125 return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2126 0, uap->amode)); 2127} 2128 2129#ifndef _SYS_SYSPROTO_H_ 2130struct faccessat_args { 2131 int dirfd; 2132 char *path; 2133 int amode; 2134 int flag; 2135} 2136#endif 2137int 2138sys_faccessat(struct thread *td, struct faccessat_args *uap) 2139{ 2140 2141 return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag, 2142 uap->amode)); 2143} 2144 2145int 2146kern_accessat(struct thread *td, int fd, const char *path, 2147 enum uio_seg pathseg, int flag, int amode) 2148{ 2149 struct ucred *cred, *usecred; 2150 struct vnode *vp; 2151 struct nameidata nd; 2152 int error; 2153 2154 if ((flag & ~(AT_EACCESS | AT_RESOLVE_BENEATH | AT_EMPTY_PATH)) != 0) 2155 return (EINVAL); 2156 if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0) 2157 return (EINVAL); 2158 2159 /* 2160 * Create and modify a temporary credential instead of one that 2161 * is potentially shared (if we need one). 2162 */ 2163 cred = td->td_ucred; 2164 if ((flag & AT_EACCESS) == 0 && 2165 ((cred->cr_uid != cred->cr_ruid || 2166 cred->cr_rgid != cred->cr_groups[0]))) { 2167 usecred = crdup(cred); 2168 usecred->cr_uid = cred->cr_ruid; 2169 usecred->cr_groups[0] = cred->cr_rgid; 2170 td->td_ucred = usecred; 2171 } else 2172 usecred = cred; 2173 AUDIT_ARG_VALUE(amode); 2174 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | 2175 AUDITVNODE1 | at2cnpflags(flag, AT_RESOLVE_BENEATH | 2176 AT_EMPTY_PATH), pathseg, path, fd, &cap_fstat_rights); 2177 if ((error = namei(&nd)) != 0) 2178 goto out; 2179 vp = nd.ni_vp; 2180 2181 error = vn_access(vp, amode, usecred, td); 2182 NDFREE_PNBUF(&nd); 2183 vput(vp); 2184out: 2185 if (usecred != cred) { 2186 td->td_ucred = cred; 2187 crfree(usecred); 2188 } 2189 return (error); 2190} 2191 2192/* 2193 * Check access permissions using "effective" credentials. 2194 */ 2195#ifndef _SYS_SYSPROTO_H_ 2196struct eaccess_args { 2197 char *path; 2198 int amode; 2199}; 2200#endif 2201int 2202sys_eaccess(struct thread *td, struct eaccess_args *uap) 2203{ 2204 2205 return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2206 AT_EACCESS, uap->amode)); 2207} 2208 2209#if defined(COMPAT_43) 2210/* 2211 * Get file status; this version follows links. 2212 */ 2213#ifndef _SYS_SYSPROTO_H_ 2214struct ostat_args { 2215 char *path; 2216 struct ostat *ub; 2217}; 2218#endif 2219int 2220ostat(struct thread *td, struct ostat_args *uap) 2221{ 2222 struct stat sb; 2223 struct ostat osb; 2224 int error; 2225 2226 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE, &sb); 2227 if (error != 0) 2228 return (error); 2229 cvtstat(&sb, &osb); 2230 return (copyout(&osb, uap->ub, sizeof (osb))); 2231} 2232 2233/* 2234 * Get file status; this version does not follow links. 2235 */ 2236#ifndef _SYS_SYSPROTO_H_ 2237struct olstat_args { 2238 char *path; 2239 struct ostat *ub; 2240}; 2241#endif 2242int 2243olstat(struct thread *td, struct olstat_args *uap) 2244{ 2245 struct stat sb; 2246 struct ostat osb; 2247 int error; 2248 2249 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path, 2250 UIO_USERSPACE, &sb); 2251 if (error != 0) 2252 return (error); 2253 cvtstat(&sb, &osb); 2254 return (copyout(&osb, uap->ub, sizeof (osb))); 2255} 2256 2257/* 2258 * Convert from an old to a new stat structure. 2259 * XXX: many values are blindly truncated. 2260 */ 2261void 2262cvtstat(struct stat *st, struct ostat *ost) 2263{ 2264 2265 bzero(ost, sizeof(*ost)); 2266 ost->st_dev = st->st_dev; 2267 ost->st_ino = st->st_ino; 2268 ost->st_mode = st->st_mode; 2269 ost->st_nlink = st->st_nlink; 2270 ost->st_uid = st->st_uid; 2271 ost->st_gid = st->st_gid; 2272 ost->st_rdev = st->st_rdev; 2273 ost->st_size = MIN(st->st_size, INT32_MAX); 2274 ost->st_atim = st->st_atim; 2275 ost->st_mtim = st->st_mtim; 2276 ost->st_ctim = st->st_ctim; 2277 ost->st_blksize = st->st_blksize; 2278 ost->st_blocks = st->st_blocks; 2279 ost->st_flags = st->st_flags; 2280 ost->st_gen = st->st_gen; 2281} 2282#endif /* COMPAT_43 */ 2283 2284#if defined(COMPAT_43) || defined(COMPAT_FREEBSD11) 2285int ino64_trunc_error; 2286SYSCTL_INT(_vfs, OID_AUTO, ino64_trunc_error, CTLFLAG_RW, 2287 &ino64_trunc_error, 0, 2288 "Error on truncation of device, file or inode number, or link count"); 2289 2290int 2291freebsd11_cvtstat(struct stat *st, struct freebsd11_stat *ost) 2292{ 2293 2294 ost->st_dev = st->st_dev; 2295 if (ost->st_dev != st->st_dev) { 2296 switch (ino64_trunc_error) { 2297 default: 2298 /* 2299 * Since dev_t is almost raw, don't clamp to the 2300 * maximum for case 2, but ignore the error. 2301 */ 2302 break; 2303 case 1: 2304 return (EOVERFLOW); 2305 } 2306 } 2307 ost->st_ino = st->st_ino; 2308 if (ost->st_ino != st->st_ino) { 2309 switch (ino64_trunc_error) { 2310 default: 2311 case 0: 2312 break; 2313 case 1: 2314 return (EOVERFLOW); 2315 case 2: 2316 ost->st_ino = UINT32_MAX; 2317 break; 2318 } 2319 } 2320 ost->st_mode = st->st_mode; 2321 ost->st_nlink = st->st_nlink; 2322 if (ost->st_nlink != st->st_nlink) { 2323 switch (ino64_trunc_error) { 2324 default: 2325 case 0: 2326 break; 2327 case 1: 2328 return (EOVERFLOW); 2329 case 2: 2330 ost->st_nlink = UINT16_MAX; 2331 break; 2332 } 2333 } 2334 ost->st_uid = st->st_uid; 2335 ost->st_gid = st->st_gid; 2336 ost->st_rdev = st->st_rdev; 2337 if (ost->st_rdev != st->st_rdev) { 2338 switch (ino64_trunc_error) { 2339 default: 2340 break; 2341 case 1: 2342 return (EOVERFLOW); 2343 } 2344 } 2345 ost->st_atim = st->st_atim; 2346 ost->st_mtim = st->st_mtim; 2347 ost->st_ctim = st->st_ctim; 2348 ost->st_size = st->st_size; 2349 ost->st_blocks = st->st_blocks; 2350 ost->st_blksize = st->st_blksize; 2351 ost->st_flags = st->st_flags; 2352 ost->st_gen = st->st_gen; 2353 ost->st_lspare = 0; 2354 ost->st_birthtim = st->st_birthtim; 2355 bzero((char *)&ost->st_birthtim + sizeof(ost->st_birthtim), 2356 sizeof(*ost) - offsetof(struct freebsd11_stat, 2357 st_birthtim) - sizeof(ost->st_birthtim)); 2358 return (0); 2359} 2360 2361int 2362freebsd11_stat(struct thread *td, struct freebsd11_stat_args* uap) 2363{ 2364 struct stat sb; 2365 struct freebsd11_stat osb; 2366 int error; 2367 2368 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE, &sb); 2369 if (error != 0) 2370 return (error); 2371 error = freebsd11_cvtstat(&sb, &osb); 2372 if (error == 0) 2373 error = copyout(&osb, uap->ub, sizeof(osb)); 2374 return (error); 2375} 2376 2377int 2378freebsd11_lstat(struct thread *td, struct freebsd11_lstat_args* uap) 2379{ 2380 struct stat sb; 2381 struct freebsd11_stat osb; 2382 int error; 2383 2384 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path, 2385 UIO_USERSPACE, &sb); 2386 if (error != 0) 2387 return (error); 2388 error = freebsd11_cvtstat(&sb, &osb); 2389 if (error == 0) 2390 error = copyout(&osb, uap->ub, sizeof(osb)); 2391 return (error); 2392} 2393 2394int 2395freebsd11_fhstat(struct thread *td, struct freebsd11_fhstat_args* uap) 2396{ 2397 struct fhandle fh; 2398 struct stat sb; 2399 struct freebsd11_stat osb; 2400 int error; 2401 2402 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t)); 2403 if (error != 0) 2404 return (error); 2405 error = kern_fhstat(td, fh, &sb); 2406 if (error != 0) 2407 return (error); 2408 error = freebsd11_cvtstat(&sb, &osb); 2409 if (error == 0) 2410 error = copyout(&osb, uap->sb, sizeof(osb)); 2411 return (error); 2412} 2413 2414int 2415freebsd11_fstatat(struct thread *td, struct freebsd11_fstatat_args* uap) 2416{ 2417 struct stat sb; 2418 struct freebsd11_stat osb; 2419 int error; 2420 2421 error = kern_statat(td, uap->flag, uap->fd, uap->path, 2422 UIO_USERSPACE, &sb); 2423 if (error != 0) 2424 return (error); 2425 error = freebsd11_cvtstat(&sb, &osb); 2426 if (error == 0) 2427 error = copyout(&osb, uap->buf, sizeof(osb)); 2428 return (error); 2429} 2430#endif /* COMPAT_FREEBSD11 */ 2431 2432/* 2433 * Get file status 2434 */ 2435#ifndef _SYS_SYSPROTO_H_ 2436struct fstatat_args { 2437 int fd; 2438 char *path; 2439 struct stat *buf; 2440 int flag; 2441} 2442#endif 2443int 2444sys_fstatat(struct thread *td, struct fstatat_args *uap) 2445{ 2446 struct stat sb; 2447 int error; 2448 2449 error = kern_statat(td, uap->flag, uap->fd, uap->path, 2450 UIO_USERSPACE, &sb); 2451 if (error == 0) 2452 error = copyout(&sb, uap->buf, sizeof (sb)); 2453 return (error); 2454} 2455 2456int 2457kern_statat(struct thread *td, int flag, int fd, const char *path, 2458 enum uio_seg pathseg, struct stat *sbp) 2459{ 2460 struct nameidata nd; 2461 int error; 2462 2463 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH | 2464 AT_EMPTY_PATH)) != 0) 2465 return (EINVAL); 2466 2467 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_RESOLVE_BENEATH | 2468 AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH) | LOCKSHARED | LOCKLEAF | 2469 AUDITVNODE1, pathseg, path, fd, &cap_fstat_rights); 2470 2471 if ((error = namei(&nd)) != 0) { 2472 if (error == ENOTDIR && 2473 (nd.ni_resflags & NIRES_EMPTYPATH) != 0) 2474 error = kern_fstat(td, fd, sbp); 2475 return (error); 2476 } 2477 error = VOP_STAT(nd.ni_vp, sbp, td->td_ucred, NOCRED); 2478 NDFREE_PNBUF(&nd); 2479 vput(nd.ni_vp); 2480#ifdef __STAT_TIME_T_EXT 2481 sbp->st_atim_ext = 0; 2482 sbp->st_mtim_ext = 0; 2483 sbp->st_ctim_ext = 0; 2484 sbp->st_btim_ext = 0; 2485#endif 2486#ifdef KTRACE 2487 if (KTRPOINT(td, KTR_STRUCT)) 2488 ktrstat_error(sbp, error); 2489#endif 2490 return (error); 2491} 2492 2493#if defined(COMPAT_FREEBSD11) 2494/* 2495 * Implementation of the NetBSD [l]stat() functions. 2496 */ 2497int 2498freebsd11_cvtnstat(struct stat *sb, struct nstat *nsb) 2499{ 2500 struct freebsd11_stat sb11; 2501 int error; 2502 2503 error = freebsd11_cvtstat(sb, &sb11); 2504 if (error != 0) 2505 return (error); 2506 2507 bzero(nsb, sizeof(*nsb)); 2508 CP(sb11, *nsb, st_dev); 2509 CP(sb11, *nsb, st_ino); 2510 CP(sb11, *nsb, st_mode); 2511 CP(sb11, *nsb, st_nlink); 2512 CP(sb11, *nsb, st_uid); 2513 CP(sb11, *nsb, st_gid); 2514 CP(sb11, *nsb, st_rdev); 2515 CP(sb11, *nsb, st_atim); 2516 CP(sb11, *nsb, st_mtim); 2517 CP(sb11, *nsb, st_ctim); 2518 CP(sb11, *nsb, st_size); 2519 CP(sb11, *nsb, st_blocks); 2520 CP(sb11, *nsb, st_blksize); 2521 CP(sb11, *nsb, st_flags); 2522 CP(sb11, *nsb, st_gen); 2523 CP(sb11, *nsb, st_birthtim); 2524 return (0); 2525} 2526 2527#ifndef _SYS_SYSPROTO_H_ 2528struct freebsd11_nstat_args { 2529 char *path; 2530 struct nstat *ub; 2531}; 2532#endif 2533int 2534freebsd11_nstat(struct thread *td, struct freebsd11_nstat_args *uap) 2535{ 2536 struct stat sb; 2537 struct nstat nsb; 2538 int error; 2539 2540 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE, &sb); 2541 if (error != 0) 2542 return (error); 2543 error = freebsd11_cvtnstat(&sb, &nsb); 2544 if (error == 0) 2545 error = copyout(&nsb, uap->ub, sizeof (nsb)); 2546 return (error); 2547} 2548 2549/* 2550 * NetBSD lstat. Get file status; this version does not follow links. 2551 */ 2552#ifndef _SYS_SYSPROTO_H_ 2553struct freebsd11_nlstat_args { 2554 char *path; 2555 struct nstat *ub; 2556}; 2557#endif 2558int 2559freebsd11_nlstat(struct thread *td, struct freebsd11_nlstat_args *uap) 2560{ 2561 struct stat sb; 2562 struct nstat nsb; 2563 int error; 2564 2565 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path, 2566 UIO_USERSPACE, &sb); 2567 if (error != 0) 2568 return (error); 2569 error = freebsd11_cvtnstat(&sb, &nsb); 2570 if (error == 0) 2571 error = copyout(&nsb, uap->ub, sizeof (nsb)); 2572 return (error); 2573} 2574#endif /* COMPAT_FREEBSD11 */ 2575 2576/* 2577 * Get configurable pathname variables. 2578 */ 2579#ifndef _SYS_SYSPROTO_H_ 2580struct pathconf_args { 2581 char *path; 2582 int name; 2583}; 2584#endif 2585int 2586sys_pathconf(struct thread *td, struct pathconf_args *uap) 2587{ 2588 long value; 2589 int error; 2590 2591 error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW, 2592 &value); 2593 if (error == 0) 2594 td->td_retval[0] = value; 2595 return (error); 2596} 2597 2598#ifndef _SYS_SYSPROTO_H_ 2599struct lpathconf_args { 2600 char *path; 2601 int name; 2602}; 2603#endif 2604int 2605sys_lpathconf(struct thread *td, struct lpathconf_args *uap) 2606{ 2607 long value; 2608 int error; 2609 2610 error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, 2611 NOFOLLOW, &value); 2612 if (error == 0) 2613 td->td_retval[0] = value; 2614 return (error); 2615} 2616 2617int 2618kern_pathconf(struct thread *td, const char *path, enum uio_seg pathseg, 2619 int name, u_long flags, long *valuep) 2620{ 2621 struct nameidata nd; 2622 int error; 2623 2624 NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags, 2625 pathseg, path); 2626 if ((error = namei(&nd)) != 0) 2627 return (error); 2628 NDFREE_PNBUF(&nd); 2629 2630 error = VOP_PATHCONF(nd.ni_vp, name, valuep); 2631 vput(nd.ni_vp); 2632 return (error); 2633} 2634 2635/* 2636 * Return target name of a symbolic link. 2637 */ 2638#ifndef _SYS_SYSPROTO_H_ 2639struct readlink_args { 2640 char *path; 2641 char *buf; 2642 size_t count; 2643}; 2644#endif 2645int 2646sys_readlink(struct thread *td, struct readlink_args *uap) 2647{ 2648 2649 return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2650 uap->buf, UIO_USERSPACE, uap->count)); 2651} 2652#ifndef _SYS_SYSPROTO_H_ 2653struct readlinkat_args { 2654 int fd; 2655 char *path; 2656 char *buf; 2657 size_t bufsize; 2658}; 2659#endif 2660int 2661sys_readlinkat(struct thread *td, struct readlinkat_args *uap) 2662{ 2663 2664 return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE, 2665 uap->buf, UIO_USERSPACE, uap->bufsize)); 2666} 2667 2668int 2669kern_readlinkat(struct thread *td, int fd, const char *path, 2670 enum uio_seg pathseg, char *buf, enum uio_seg bufseg, size_t count) 2671{ 2672 struct vnode *vp; 2673 struct nameidata nd; 2674 int error; 2675 2676 if (count > IOSIZE_MAX) 2677 return (EINVAL); 2678 2679 NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1 | 2680 EMPTYPATH, pathseg, path, fd); 2681 2682 if ((error = namei(&nd)) != 0) 2683 return (error); 2684 NDFREE_PNBUF(&nd); 2685 vp = nd.ni_vp; 2686 2687 error = kern_readlink_vp(vp, buf, bufseg, count, td); 2688 vput(vp); 2689 2690 return (error); 2691} 2692 2693/* 2694 * Helper function to readlink from a vnode 2695 */ 2696static int 2697kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg, size_t count, 2698 struct thread *td) 2699{ 2700 struct iovec aiov; 2701 struct uio auio; 2702 int error; 2703 2704 ASSERT_VOP_LOCKED(vp, "kern_readlink_vp(): vp not locked"); 2705#ifdef MAC 2706 error = mac_vnode_check_readlink(td->td_ucred, vp); 2707 if (error != 0) 2708 return (error); 2709#endif 2710 if (vp->v_type != VLNK && (vp->v_vflag & VV_READLINK) == 0) 2711 return (EINVAL); 2712 2713 aiov.iov_base = buf; 2714 aiov.iov_len = count; 2715 auio.uio_iov = &aiov; 2716 auio.uio_iovcnt = 1; 2717 auio.uio_offset = 0; 2718 auio.uio_rw = UIO_READ; 2719 auio.uio_segflg = bufseg; 2720 auio.uio_td = td; 2721 auio.uio_resid = count; 2722 error = VOP_READLINK(vp, &auio, td->td_ucred); 2723 td->td_retval[0] = count - auio.uio_resid; 2724 return (error); 2725} 2726 2727/* 2728 * Common implementation code for chflags() and fchflags(). 2729 */ 2730static int 2731setfflags(struct thread *td, struct vnode *vp, u_long flags) 2732{ 2733 struct mount *mp; 2734 struct vattr vattr; 2735 int error; 2736 2737 /* We can't support the value matching VNOVAL. */ 2738 if (flags == VNOVAL) 2739 return (EOPNOTSUPP); 2740 2741 /* 2742 * Prevent non-root users from setting flags on devices. When 2743 * a device is reused, users can retain ownership of the device 2744 * if they are allowed to set flags and programs assume that 2745 * chown can't fail when done as root. 2746 */ 2747 if (vp->v_type == VCHR || vp->v_type == VBLK) { 2748 error = priv_check(td, PRIV_VFS_CHFLAGS_DEV); 2749 if (error != 0) 2750 return (error); 2751 } 2752 2753 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2754 return (error); 2755 VATTR_NULL(&vattr); 2756 vattr.va_flags = flags; 2757 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2758#ifdef MAC 2759 error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags); 2760 if (error == 0) 2761#endif 2762 error = VOP_SETATTR(vp, &vattr, td->td_ucred); 2763 VOP_UNLOCK(vp); 2764 vn_finished_write(mp); 2765 return (error); 2766} 2767 2768/* 2769 * Change flags of a file given a path name. 2770 */ 2771#ifndef _SYS_SYSPROTO_H_ 2772struct chflags_args { 2773 const char *path; 2774 u_long flags; 2775}; 2776#endif 2777int 2778sys_chflags(struct thread *td, struct chflags_args *uap) 2779{ 2780 2781 return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2782 uap->flags, 0)); 2783} 2784 2785#ifndef _SYS_SYSPROTO_H_ 2786struct chflagsat_args { 2787 int fd; 2788 const char *path; 2789 u_long flags; 2790 int atflag; 2791} 2792#endif 2793int 2794sys_chflagsat(struct thread *td, struct chflagsat_args *uap) 2795{ 2796 2797 return (kern_chflagsat(td, uap->fd, uap->path, UIO_USERSPACE, 2798 uap->flags, uap->atflag)); 2799} 2800 2801/* 2802 * Same as chflags() but doesn't follow symlinks. 2803 */ 2804#ifndef _SYS_SYSPROTO_H_ 2805struct lchflags_args { 2806 const char *path; 2807 u_long flags; 2808}; 2809#endif 2810int 2811sys_lchflags(struct thread *td, struct lchflags_args *uap) 2812{ 2813 2814 return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2815 uap->flags, AT_SYMLINK_NOFOLLOW)); 2816} 2817 2818static int 2819kern_chflagsat(struct thread *td, int fd, const char *path, 2820 enum uio_seg pathseg, u_long flags, int atflag) 2821{ 2822 struct nameidata nd; 2823 int error; 2824 2825 if ((atflag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH | 2826 AT_EMPTY_PATH)) != 0) 2827 return (EINVAL); 2828 2829 AUDIT_ARG_FFLAGS(flags); 2830 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(atflag, AT_SYMLINK_NOFOLLOW | 2831 AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path, 2832 fd, &cap_fchflags_rights); 2833 if ((error = namei(&nd)) != 0) 2834 return (error); 2835 NDFREE_PNBUF(&nd); 2836 error = setfflags(td, nd.ni_vp, flags); 2837 vrele(nd.ni_vp); 2838 return (error); 2839} 2840 2841/* 2842 * Change flags of a file given a file descriptor. 2843 */ 2844#ifndef _SYS_SYSPROTO_H_ 2845struct fchflags_args { 2846 int fd; 2847 u_long flags; 2848}; 2849#endif 2850int 2851sys_fchflags(struct thread *td, struct fchflags_args *uap) 2852{ 2853 struct file *fp; 2854 int error; 2855 2856 AUDIT_ARG_FD(uap->fd); 2857 AUDIT_ARG_FFLAGS(uap->flags); 2858 error = getvnode(td, uap->fd, &cap_fchflags_rights, 2859 &fp); 2860 if (error != 0) 2861 return (error); 2862#ifdef AUDIT 2863 if (AUDITING_TD(td)) { 2864 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY); 2865 AUDIT_ARG_VNODE1(fp->f_vnode); 2866 VOP_UNLOCK(fp->f_vnode); 2867 } 2868#endif 2869 error = setfflags(td, fp->f_vnode, uap->flags); 2870 fdrop(fp, td); 2871 return (error); 2872} 2873 2874/* 2875 * Common implementation code for chmod(), lchmod() and fchmod(). 2876 */ 2877int 2878setfmode(struct thread *td, struct ucred *cred, struct vnode *vp, int mode) 2879{ 2880 struct mount *mp; 2881 struct vattr vattr; 2882 int error; 2883 2884 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2885 return (error); 2886 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2887 VATTR_NULL(&vattr); 2888 vattr.va_mode = mode & ALLPERMS; 2889#ifdef MAC 2890 error = mac_vnode_check_setmode(cred, vp, vattr.va_mode); 2891 if (error == 0) 2892#endif 2893 error = VOP_SETATTR(vp, &vattr, cred); 2894 VOP_UNLOCK(vp); 2895 vn_finished_write(mp); 2896 return (error); 2897} 2898 2899/* 2900 * Change mode of a file given path name. 2901 */ 2902#ifndef _SYS_SYSPROTO_H_ 2903struct chmod_args { 2904 char *path; 2905 int mode; 2906}; 2907#endif 2908int 2909sys_chmod(struct thread *td, struct chmod_args *uap) 2910{ 2911 2912 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2913 uap->mode, 0)); 2914} 2915 2916#ifndef _SYS_SYSPROTO_H_ 2917struct fchmodat_args { 2918 int dirfd; 2919 char *path; 2920 mode_t mode; 2921 int flag; 2922} 2923#endif 2924int 2925sys_fchmodat(struct thread *td, struct fchmodat_args *uap) 2926{ 2927 2928 return (kern_fchmodat(td, uap->fd, uap->path, UIO_USERSPACE, 2929 uap->mode, uap->flag)); 2930} 2931 2932/* 2933 * Change mode of a file given path name (don't follow links.) 2934 */ 2935#ifndef _SYS_SYSPROTO_H_ 2936struct lchmod_args { 2937 char *path; 2938 int mode; 2939}; 2940#endif 2941int 2942sys_lchmod(struct thread *td, struct lchmod_args *uap) 2943{ 2944 2945 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2946 uap->mode, AT_SYMLINK_NOFOLLOW)); 2947} 2948 2949int 2950kern_fchmodat(struct thread *td, int fd, const char *path, 2951 enum uio_seg pathseg, mode_t mode, int flag) 2952{ 2953 struct nameidata nd; 2954 int error; 2955 2956 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH | 2957 AT_EMPTY_PATH)) != 0) 2958 return (EINVAL); 2959 2960 AUDIT_ARG_MODE(mode); 2961 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW | 2962 AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path, 2963 fd, &cap_fchmod_rights); 2964 if ((error = namei(&nd)) != 0) 2965 return (error); 2966 NDFREE_PNBUF(&nd); 2967 error = setfmode(td, td->td_ucred, nd.ni_vp, mode); 2968 vrele(nd.ni_vp); 2969 return (error); 2970} 2971 2972/* 2973 * Change mode of a file given a file descriptor. 2974 */ 2975#ifndef _SYS_SYSPROTO_H_ 2976struct fchmod_args { 2977 int fd; 2978 int mode; 2979}; 2980#endif 2981int 2982sys_fchmod(struct thread *td, struct fchmod_args *uap) 2983{ 2984 struct file *fp; 2985 int error; 2986 2987 AUDIT_ARG_FD(uap->fd); 2988 AUDIT_ARG_MODE(uap->mode); 2989 2990 error = fget(td, uap->fd, &cap_fchmod_rights, &fp); 2991 if (error != 0) 2992 return (error); 2993 error = fo_chmod(fp, uap->mode, td->td_ucred, td); 2994 fdrop(fp, td); 2995 return (error); 2996} 2997 2998/* 2999 * Common implementation for chown(), lchown(), and fchown() 3000 */ 3001int 3002setfown(struct thread *td, struct ucred *cred, struct vnode *vp, uid_t uid, 3003 gid_t gid) 3004{ 3005 struct mount *mp; 3006 struct vattr vattr; 3007 int error; 3008 3009 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 3010 return (error); 3011 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3012 VATTR_NULL(&vattr); 3013 vattr.va_uid = uid; 3014 vattr.va_gid = gid; 3015#ifdef MAC 3016 error = mac_vnode_check_setowner(cred, vp, vattr.va_uid, 3017 vattr.va_gid); 3018 if (error == 0) 3019#endif 3020 error = VOP_SETATTR(vp, &vattr, cred); 3021 VOP_UNLOCK(vp); 3022 vn_finished_write(mp); 3023 return (error); 3024} 3025 3026/* 3027 * Set ownership given a path name. 3028 */ 3029#ifndef _SYS_SYSPROTO_H_ 3030struct chown_args { 3031 char *path; 3032 int uid; 3033 int gid; 3034}; 3035#endif 3036int 3037sys_chown(struct thread *td, struct chown_args *uap) 3038{ 3039 3040 return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid, 3041 uap->gid, 0)); 3042} 3043 3044#ifndef _SYS_SYSPROTO_H_ 3045struct fchownat_args { 3046 int fd; 3047 const char * path; 3048 uid_t uid; 3049 gid_t gid; 3050 int flag; 3051}; 3052#endif 3053int 3054sys_fchownat(struct thread *td, struct fchownat_args *uap) 3055{ 3056 3057 return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid, 3058 uap->gid, uap->flag)); 3059} 3060 3061int 3062kern_fchownat(struct thread *td, int fd, const char *path, 3063 enum uio_seg pathseg, int uid, int gid, int flag) 3064{ 3065 struct nameidata nd; 3066 int error; 3067 3068 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH | 3069 AT_EMPTY_PATH)) != 0) 3070 return (EINVAL); 3071 3072 AUDIT_ARG_OWNER(uid, gid); 3073 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW | 3074 AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path, 3075 fd, &cap_fchown_rights); 3076 3077 if ((error = namei(&nd)) != 0) 3078 return (error); 3079 NDFREE_PNBUF(&nd); 3080 error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid); 3081 vrele(nd.ni_vp); 3082 return (error); 3083} 3084 3085/* 3086 * Set ownership given a path name, do not cross symlinks. 3087 */ 3088#ifndef _SYS_SYSPROTO_H_ 3089struct lchown_args { 3090 char *path; 3091 int uid; 3092 int gid; 3093}; 3094#endif 3095int 3096sys_lchown(struct thread *td, struct lchown_args *uap) 3097{ 3098 3099 return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 3100 uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW)); 3101} 3102 3103/* 3104 * Set ownership given a file descriptor. 3105 */ 3106#ifndef _SYS_SYSPROTO_H_ 3107struct fchown_args { 3108 int fd; 3109 int uid; 3110 int gid; 3111}; 3112#endif 3113int 3114sys_fchown(struct thread *td, struct fchown_args *uap) 3115{ 3116 struct file *fp; 3117 int error; 3118 3119 AUDIT_ARG_FD(uap->fd); 3120 AUDIT_ARG_OWNER(uap->uid, uap->gid); 3121 error = fget(td, uap->fd, &cap_fchown_rights, &fp); 3122 if (error != 0) 3123 return (error); 3124 error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td); 3125 fdrop(fp, td); 3126 return (error); 3127} 3128 3129/* 3130 * Common implementation code for utimes(), lutimes(), and futimes(). 3131 */ 3132static int 3133getutimes(const struct timeval *usrtvp, enum uio_seg tvpseg, 3134 struct timespec *tsp) 3135{ 3136 struct timeval tv[2]; 3137 const struct timeval *tvp; 3138 int error; 3139 3140 if (usrtvp == NULL) { 3141 vfs_timestamp(&tsp[0]); 3142 tsp[1] = tsp[0]; 3143 } else { 3144 if (tvpseg == UIO_SYSSPACE) { 3145 tvp = usrtvp; 3146 } else { 3147 if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0) 3148 return (error); 3149 tvp = tv; 3150 } 3151 3152 if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 || 3153 tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000) 3154 return (EINVAL); 3155 TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]); 3156 TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]); 3157 } 3158 return (0); 3159} 3160 3161/* 3162 * Common implementation code for futimens(), utimensat(). 3163 */ 3164#define UTIMENS_NULL 0x1 3165#define UTIMENS_EXIT 0x2 3166static int 3167getutimens(const struct timespec *usrtsp, enum uio_seg tspseg, 3168 struct timespec *tsp, int *retflags) 3169{ 3170 struct timespec tsnow; 3171 int error; 3172 3173 vfs_timestamp(&tsnow); 3174 *retflags = 0; 3175 if (usrtsp == NULL) { 3176 tsp[0] = tsnow; 3177 tsp[1] = tsnow; 3178 *retflags |= UTIMENS_NULL; 3179 return (0); 3180 } 3181 if (tspseg == UIO_SYSSPACE) { 3182 tsp[0] = usrtsp[0]; 3183 tsp[1] = usrtsp[1]; 3184 } else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0) 3185 return (error); 3186 if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT) 3187 *retflags |= UTIMENS_EXIT; 3188 if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW) 3189 *retflags |= UTIMENS_NULL; 3190 if (tsp[0].tv_nsec == UTIME_OMIT) 3191 tsp[0].tv_sec = VNOVAL; 3192 else if (tsp[0].tv_nsec == UTIME_NOW) 3193 tsp[0] = tsnow; 3194 else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L) 3195 return (EINVAL); 3196 if (tsp[1].tv_nsec == UTIME_OMIT) 3197 tsp[1].tv_sec = VNOVAL; 3198 else if (tsp[1].tv_nsec == UTIME_NOW) 3199 tsp[1] = tsnow; 3200 else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L) 3201 return (EINVAL); 3202 3203 return (0); 3204} 3205 3206/* 3207 * Common implementation code for utimes(), lutimes(), futimes(), futimens(), 3208 * and utimensat(). 3209 */ 3210static int 3211setutimes(struct thread *td, struct vnode *vp, const struct timespec *ts, 3212 int numtimes, int nullflag) 3213{ 3214 struct mount *mp; 3215 struct vattr vattr; 3216 int error; 3217 bool setbirthtime; 3218 3219 setbirthtime = false; 3220 vattr.va_birthtime.tv_sec = VNOVAL; 3221 vattr.va_birthtime.tv_nsec = 0; 3222 3223 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 3224 return (error); 3225 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3226 if (numtimes < 3 && VOP_GETATTR(vp, &vattr, td->td_ucred) == 0 && 3227 timespeccmp(&ts[1], &vattr.va_birthtime, < )) 3228 setbirthtime = true; 3229 VATTR_NULL(&vattr); 3230 vattr.va_atime = ts[0]; 3231 vattr.va_mtime = ts[1]; 3232 if (setbirthtime) 3233 vattr.va_birthtime = ts[1]; 3234 if (numtimes > 2) 3235 vattr.va_birthtime = ts[2]; 3236 if (nullflag) 3237 vattr.va_vaflags |= VA_UTIMES_NULL; 3238#ifdef MAC 3239 error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime, 3240 vattr.va_mtime); 3241#endif 3242 if (error == 0) 3243 error = VOP_SETATTR(vp, &vattr, td->td_ucred); 3244 VOP_UNLOCK(vp); 3245 vn_finished_write(mp); 3246 return (error); 3247} 3248 3249/* 3250 * Set the access and modification times of a file. 3251 */ 3252#ifndef _SYS_SYSPROTO_H_ 3253struct utimes_args { 3254 char *path; 3255 struct timeval *tptr; 3256}; 3257#endif 3258int 3259sys_utimes(struct thread *td, struct utimes_args *uap) 3260{ 3261 3262 return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 3263 uap->tptr, UIO_USERSPACE)); 3264} 3265 3266#ifndef _SYS_SYSPROTO_H_ 3267struct futimesat_args { 3268 int fd; 3269 const char * path; 3270 const struct timeval * times; 3271}; 3272#endif 3273int 3274sys_futimesat(struct thread *td, struct futimesat_args *uap) 3275{ 3276 3277 return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE, 3278 uap->times, UIO_USERSPACE)); 3279} 3280 3281int 3282kern_utimesat(struct thread *td, int fd, const char *path, 3283 enum uio_seg pathseg, const struct timeval *tptr, enum uio_seg tptrseg) 3284{ 3285 struct nameidata nd; 3286 struct timespec ts[2]; 3287 int error; 3288 3289 if ((error = getutimes(tptr, tptrseg, ts)) != 0) 3290 return (error); 3291 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd, 3292 &cap_futimes_rights); 3293 3294 if ((error = namei(&nd)) != 0) 3295 return (error); 3296 NDFREE_PNBUF(&nd); 3297 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL); 3298 vrele(nd.ni_vp); 3299 return (error); 3300} 3301 3302/* 3303 * Set the access and modification times of a file. 3304 */ 3305#ifndef _SYS_SYSPROTO_H_ 3306struct lutimes_args { 3307 char *path; 3308 struct timeval *tptr; 3309}; 3310#endif 3311int 3312sys_lutimes(struct thread *td, struct lutimes_args *uap) 3313{ 3314 3315 return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr, 3316 UIO_USERSPACE)); 3317} 3318 3319int 3320kern_lutimes(struct thread *td, const char *path, enum uio_seg pathseg, 3321 const struct timeval *tptr, enum uio_seg tptrseg) 3322{ 3323 struct timespec ts[2]; 3324 struct nameidata nd; 3325 int error; 3326 3327 if ((error = getutimes(tptr, tptrseg, ts)) != 0) 3328 return (error); 3329 NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path); 3330 if ((error = namei(&nd)) != 0) 3331 return (error); 3332 NDFREE_PNBUF(&nd); 3333 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL); 3334 vrele(nd.ni_vp); 3335 return (error); 3336} 3337 3338/* 3339 * Set the access and modification times of a file. 3340 */ 3341#ifndef _SYS_SYSPROTO_H_ 3342struct futimes_args { 3343 int fd; 3344 struct timeval *tptr; 3345}; 3346#endif 3347int 3348sys_futimes(struct thread *td, struct futimes_args *uap) 3349{ 3350 3351 return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE)); 3352} 3353 3354int 3355kern_futimes(struct thread *td, int fd, const struct timeval *tptr, 3356 enum uio_seg tptrseg) 3357{ 3358 struct timespec ts[2]; 3359 struct file *fp; 3360 int error; 3361 3362 AUDIT_ARG_FD(fd); 3363 error = getutimes(tptr, tptrseg, ts); 3364 if (error != 0) 3365 return (error); 3366 error = getvnode(td, fd, &cap_futimes_rights, &fp); 3367 if (error != 0) 3368 return (error); 3369#ifdef AUDIT 3370 if (AUDITING_TD(td)) { 3371 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY); 3372 AUDIT_ARG_VNODE1(fp->f_vnode); 3373 VOP_UNLOCK(fp->f_vnode); 3374 } 3375#endif 3376 error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL); 3377 fdrop(fp, td); 3378 return (error); 3379} 3380 3381int 3382sys_futimens(struct thread *td, struct futimens_args *uap) 3383{ 3384 3385 return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE)); 3386} 3387 3388int 3389kern_futimens(struct thread *td, int fd, const struct timespec *tptr, 3390 enum uio_seg tptrseg) 3391{ 3392 struct timespec ts[2]; 3393 struct file *fp; 3394 int error, flags; 3395 3396 AUDIT_ARG_FD(fd); 3397 error = getutimens(tptr, tptrseg, ts, &flags); 3398 if (error != 0) 3399 return (error); 3400 if (flags & UTIMENS_EXIT) 3401 return (0); 3402 error = getvnode(td, fd, &cap_futimes_rights, &fp); 3403 if (error != 0) 3404 return (error); 3405#ifdef AUDIT 3406 if (AUDITING_TD(td)) { 3407 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY); 3408 AUDIT_ARG_VNODE1(fp->f_vnode); 3409 VOP_UNLOCK(fp->f_vnode); 3410 } 3411#endif 3412 error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL); 3413 fdrop(fp, td); 3414 return (error); 3415} 3416 3417int 3418sys_utimensat(struct thread *td, struct utimensat_args *uap) 3419{ 3420 3421 return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE, 3422 uap->times, UIO_USERSPACE, uap->flag)); 3423} 3424 3425int 3426kern_utimensat(struct thread *td, int fd, const char *path, 3427 enum uio_seg pathseg, const struct timespec *tptr, enum uio_seg tptrseg, 3428 int flag) 3429{ 3430 struct nameidata nd; 3431 struct timespec ts[2]; 3432 int error, flags; 3433 3434 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH | 3435 AT_EMPTY_PATH)) != 0) 3436 return (EINVAL); 3437 3438 if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0) 3439 return (error); 3440 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW | 3441 AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, 3442 pathseg, path, fd, &cap_futimes_rights); 3443 if ((error = namei(&nd)) != 0) 3444 return (error); 3445 /* 3446 * We are allowed to call namei() regardless of 2xUTIME_OMIT. 3447 * POSIX states: 3448 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected." 3449 * "Search permission is denied by a component of the path prefix." 3450 */ 3451 NDFREE_PNBUF(&nd); 3452 if ((flags & UTIMENS_EXIT) == 0) 3453 error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL); 3454 vrele(nd.ni_vp); 3455 return (error); 3456} 3457 3458/* 3459 * Truncate a file given its path name. 3460 */ 3461#ifndef _SYS_SYSPROTO_H_ 3462struct truncate_args { 3463 char *path; 3464 int pad; 3465 off_t length; 3466}; 3467#endif 3468int 3469sys_truncate(struct thread *td, struct truncate_args *uap) 3470{ 3471 3472 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length)); 3473} 3474 3475int 3476kern_truncate(struct thread *td, const char *path, enum uio_seg pathseg, 3477 off_t length) 3478{ 3479 struct mount *mp; 3480 struct vnode *vp; 3481 void *rl_cookie; 3482 struct nameidata nd; 3483 int error; 3484 3485 if (length < 0) 3486 return (EINVAL); 3487 NDPREINIT(&nd); 3488retry: 3489 NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path); 3490 if ((error = namei(&nd)) != 0) 3491 return (error); 3492 vp = nd.ni_vp; 3493 NDFREE_PNBUF(&nd); 3494 rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX); 3495 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3496 vn_rangelock_unlock(vp, rl_cookie); 3497 vrele(vp); 3498 return (error); 3499 } 3500 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3501 if (vp->v_type == VDIR) { 3502 error = EISDIR; 3503 goto out; 3504 } 3505#ifdef MAC 3506 error = mac_vnode_check_write(td->td_ucred, NOCRED, vp); 3507 if (error != 0) 3508 goto out; 3509#endif 3510 error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td); 3511 if (error != 0) 3512 goto out; 3513 3514 error = vn_truncate_locked(vp, length, false, td->td_ucred); 3515out: 3516 VOP_UNLOCK(vp); 3517 vn_finished_write(mp); 3518 vn_rangelock_unlock(vp, rl_cookie); 3519 vrele(vp); 3520 if (error == ERELOOKUP) 3521 goto retry; 3522 return (error); 3523} 3524 3525#if defined(COMPAT_43) 3526/* 3527 * Truncate a file given its path name. 3528 */ 3529#ifndef _SYS_SYSPROTO_H_ 3530struct otruncate_args { 3531 char *path; 3532 long length; 3533}; 3534#endif 3535int 3536otruncate(struct thread *td, struct otruncate_args *uap) 3537{ 3538 3539 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length)); 3540} 3541#endif /* COMPAT_43 */ 3542 3543#if defined(COMPAT_FREEBSD6) 3544/* Versions with the pad argument */ 3545int 3546freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap) 3547{ 3548 3549 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length)); 3550} 3551 3552int 3553freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap) 3554{ 3555 3556 return (kern_ftruncate(td, uap->fd, uap->length)); 3557} 3558#endif 3559 3560int 3561kern_fsync(struct thread *td, int fd, bool fullsync) 3562{ 3563 struct vnode *vp; 3564 struct mount *mp; 3565 struct file *fp; 3566 int error; 3567 3568 AUDIT_ARG_FD(fd); 3569 error = getvnode(td, fd, &cap_fsync_rights, &fp); 3570 if (error != 0) 3571 return (error); 3572 vp = fp->f_vnode; 3573#if 0 3574 if (!fullsync) 3575 /* XXXKIB: compete outstanding aio writes */; 3576#endif 3577retry: 3578 error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH); 3579 if (error != 0) 3580 goto drop; 3581 vn_lock(vp, vn_lktype_write(mp, vp) | LK_RETRY); 3582 AUDIT_ARG_VNODE1(vp); 3583 vnode_pager_clean_async(vp); 3584 error = fullsync ? VOP_FSYNC(vp, MNT_WAIT, td) : VOP_FDATASYNC(vp, td); 3585 VOP_UNLOCK(vp); 3586 vn_finished_write(mp); 3587 if (error == ERELOOKUP) 3588 goto retry; 3589drop: 3590 fdrop(fp, td); 3591 return (error); 3592} 3593 3594/* 3595 * Sync an open file. 3596 */ 3597#ifndef _SYS_SYSPROTO_H_ 3598struct fsync_args { 3599 int fd; 3600}; 3601#endif 3602int 3603sys_fsync(struct thread *td, struct fsync_args *uap) 3604{ 3605 3606 return (kern_fsync(td, uap->fd, true)); 3607} 3608 3609int 3610sys_fdatasync(struct thread *td, struct fdatasync_args *uap) 3611{ 3612 3613 return (kern_fsync(td, uap->fd, false)); 3614} 3615 3616/* 3617 * Rename files. Source and destination must either both be directories, or 3618 * both not be directories. If target is a directory, it must be empty. 3619 */ 3620#ifndef _SYS_SYSPROTO_H_ 3621struct rename_args { 3622 char *from; 3623 char *to; 3624}; 3625#endif 3626int 3627sys_rename(struct thread *td, struct rename_args *uap) 3628{ 3629 3630 return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD, 3631 uap->to, UIO_USERSPACE)); 3632} 3633 3634#ifndef _SYS_SYSPROTO_H_ 3635struct renameat_args { 3636 int oldfd; 3637 char *old; 3638 int newfd; 3639 char *new; 3640}; 3641#endif 3642int 3643sys_renameat(struct thread *td, struct renameat_args *uap) 3644{ 3645 3646 return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new, 3647 UIO_USERSPACE)); 3648} 3649 3650#ifdef MAC 3651static int 3652kern_renameat_mac(struct thread *td, int oldfd, const char *old, int newfd, 3653 const char *new, enum uio_seg pathseg, struct nameidata *fromnd) 3654{ 3655 int error; 3656 3657 NDINIT_ATRIGHTS(fromnd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1, 3658 pathseg, old, oldfd, &cap_renameat_source_rights); 3659 if ((error = namei(fromnd)) != 0) 3660 return (error); 3661 error = mac_vnode_check_rename_from(td->td_ucred, fromnd->ni_dvp, 3662 fromnd->ni_vp, &fromnd->ni_cnd); 3663 VOP_UNLOCK(fromnd->ni_dvp); 3664 if (fromnd->ni_dvp != fromnd->ni_vp) 3665 VOP_UNLOCK(fromnd->ni_vp); 3666 if (error != 0) { 3667 NDFREE_PNBUF(fromnd); 3668 vrele(fromnd->ni_dvp); 3669 vrele(fromnd->ni_vp); 3670 } 3671 return (error); 3672} 3673#endif 3674 3675int 3676kern_renameat(struct thread *td, int oldfd, const char *old, int newfd, 3677 const char *new, enum uio_seg pathseg) 3678{ 3679 struct mount *mp = NULL; 3680 struct vnode *tvp, *fvp, *tdvp; 3681 struct nameidata fromnd, tond; 3682 uint64_t tondflags; 3683 int error; 3684 3685again: 3686 bwillwrite(); 3687#ifdef MAC 3688 if (mac_vnode_check_rename_from_enabled()) { 3689 error = kern_renameat_mac(td, oldfd, old, newfd, new, pathseg, 3690 &fromnd); 3691 if (error != 0) 3692 return (error); 3693 } else { 3694#endif 3695 NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | AUDITVNODE1, 3696 pathseg, old, oldfd, &cap_renameat_source_rights); 3697 if ((error = namei(&fromnd)) != 0) 3698 return (error); 3699#ifdef MAC 3700 } 3701#endif 3702 fvp = fromnd.ni_vp; 3703 tondflags = LOCKPARENT | LOCKLEAF | NOCACHE | AUDITVNODE2; 3704 if (fromnd.ni_vp->v_type == VDIR) 3705 tondflags |= WILLBEDIR; 3706 NDINIT_ATRIGHTS(&tond, RENAME, tondflags, pathseg, new, newfd, 3707 &cap_renameat_target_rights); 3708 if ((error = namei(&tond)) != 0) { 3709 /* Translate error code for rename("dir1", "dir2/."). */ 3710 if (error == EISDIR && fvp->v_type == VDIR) 3711 error = EINVAL; 3712 NDFREE_PNBUF(&fromnd); 3713 vrele(fromnd.ni_dvp); 3714 vrele(fvp); 3715 goto out1; 3716 } 3717 tdvp = tond.ni_dvp; 3718 tvp = tond.ni_vp; 3719 error = vn_start_write(fvp, &mp, V_NOWAIT); 3720 if (error != 0) { 3721 NDFREE_PNBUF(&fromnd); 3722 NDFREE_PNBUF(&tond); 3723 if (tvp != NULL) 3724 vput(tvp); 3725 if (tdvp == tvp) 3726 vrele(tdvp); 3727 else 3728 vput(tdvp); 3729 vrele(fromnd.ni_dvp); 3730 vrele(fvp); 3731 error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH); 3732 if (error != 0) 3733 return (error); 3734 goto again; 3735 } 3736 if (tvp != NULL) { 3737 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3738 error = ENOTDIR; 3739 goto out; 3740 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3741 error = EISDIR; 3742 goto out; 3743 } 3744#ifdef CAPABILITIES 3745 if (newfd != AT_FDCWD && (tond.ni_resflags & NIRES_ABS) == 0) { 3746 /* 3747 * If the target already exists we require CAP_UNLINKAT 3748 * from 'newfd', when newfd was used for the lookup. 3749 */ 3750 error = cap_check(&tond.ni_filecaps.fc_rights, 3751 &cap_unlinkat_rights); 3752 if (error != 0) 3753 goto out; 3754 } 3755#endif 3756 } 3757 if (fvp == tdvp) { 3758 error = EINVAL; 3759 goto out; 3760 } 3761 /* 3762 * If the source is the same as the destination (that is, if they 3763 * are links to the same vnode), then there is nothing to do. 3764 */ 3765 if (fvp == tvp) 3766 error = ERESTART; 3767#ifdef MAC 3768 else 3769 error = mac_vnode_check_rename_to(td->td_ucred, tdvp, 3770 tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd); 3771#endif 3772out: 3773 if (error == 0) { 3774 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3775 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3776 NDFREE_PNBUF(&fromnd); 3777 NDFREE_PNBUF(&tond); 3778 } else { 3779 NDFREE_PNBUF(&fromnd); 3780 NDFREE_PNBUF(&tond); 3781 if (tvp != NULL) 3782 vput(tvp); 3783 if (tdvp == tvp) 3784 vrele(tdvp); 3785 else 3786 vput(tdvp); 3787 vrele(fromnd.ni_dvp); 3788 vrele(fvp); 3789 } 3790 vn_finished_write(mp); 3791out1: 3792 if (error == ERESTART) 3793 return (0); 3794 if (error == ERELOOKUP) 3795 goto again; 3796 return (error); 3797} 3798 3799/* 3800 * Make a directory file. 3801 */ 3802#ifndef _SYS_SYSPROTO_H_ 3803struct mkdir_args { 3804 char *path; 3805 int mode; 3806}; 3807#endif 3808int 3809sys_mkdir(struct thread *td, struct mkdir_args *uap) 3810{ 3811 3812 return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 3813 uap->mode)); 3814} 3815 3816#ifndef _SYS_SYSPROTO_H_ 3817struct mkdirat_args { 3818 int fd; 3819 char *path; 3820 mode_t mode; 3821}; 3822#endif 3823int 3824sys_mkdirat(struct thread *td, struct mkdirat_args *uap) 3825{ 3826 3827 return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode)); 3828} 3829 3830int 3831kern_mkdirat(struct thread *td, int fd, const char *path, enum uio_seg segflg, 3832 int mode) 3833{ 3834 struct mount *mp; 3835 struct vattr vattr; 3836 struct nameidata nd; 3837 int error; 3838 3839 AUDIT_ARG_MODE(mode); 3840 NDPREINIT(&nd); 3841restart: 3842 bwillwrite(); 3843 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | AUDITVNODE1 | 3844 NC_NOMAKEENTRY | NC_KEEPPOSENTRY | FAILIFEXISTS | WILLBEDIR, 3845 segflg, path, fd, &cap_mkdirat_rights); 3846 if ((error = namei(&nd)) != 0) 3847 return (error); 3848 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3849 NDFREE_PNBUF(&nd); 3850 vput(nd.ni_dvp); 3851 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH)) != 0) 3852 return (error); 3853 goto restart; 3854 } 3855 VATTR_NULL(&vattr); 3856 vattr.va_type = VDIR; 3857 vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_pd->pd_cmask; 3858#ifdef MAC 3859 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd, 3860 &vattr); 3861 if (error != 0) 3862 goto out; 3863#endif 3864 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3865#ifdef MAC 3866out: 3867#endif 3868 NDFREE_PNBUF(&nd); 3869 VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true); 3870 vn_finished_write(mp); 3871 if (error == ERELOOKUP) 3872 goto restart; 3873 return (error); 3874} 3875 3876/* 3877 * Remove a directory file. 3878 */ 3879#ifndef _SYS_SYSPROTO_H_ 3880struct rmdir_args { 3881 char *path; 3882}; 3883#endif 3884int 3885sys_rmdir(struct thread *td, struct rmdir_args *uap) 3886{ 3887 3888 return (kern_frmdirat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE, 3889 0)); 3890} 3891 3892int 3893kern_frmdirat(struct thread *td, int dfd, const char *path, int fd, 3894 enum uio_seg pathseg, int flag) 3895{ 3896 struct mount *mp; 3897 struct vnode *vp; 3898 struct file *fp; 3899 struct nameidata nd; 3900 cap_rights_t rights; 3901 int error; 3902 3903 fp = NULL; 3904 if (fd != FD_NONE) { 3905 error = getvnode(td, fd, cap_rights_init_one(&rights, 3906 CAP_LOOKUP), &fp); 3907 if (error != 0) 3908 return (error); 3909 } 3910 3911 NDPREINIT(&nd); 3912restart: 3913 bwillwrite(); 3914 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 | 3915 at2cnpflags(flag, AT_RESOLVE_BENEATH), 3916 pathseg, path, dfd, &cap_unlinkat_rights); 3917 if ((error = namei(&nd)) != 0) 3918 goto fdout; 3919 vp = nd.ni_vp; 3920 if (vp->v_type != VDIR) { 3921 error = ENOTDIR; 3922 goto out; 3923 } 3924 /* 3925 * No rmdir "." please. 3926 */ 3927 if (nd.ni_dvp == vp) { 3928 error = EINVAL; 3929 goto out; 3930 } 3931 /* 3932 * The root of a mounted filesystem cannot be deleted. 3933 */ 3934 if (vp->v_vflag & VV_ROOT) { 3935 error = EBUSY; 3936 goto out; 3937 } 3938 3939 if (fp != NULL && fp->f_vnode != vp) { 3940 if (VN_IS_DOOMED(fp->f_vnode)) 3941 error = EBADF; 3942 else 3943 error = EDEADLK; 3944 goto out; 3945 } 3946 3947#ifdef MAC 3948 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp, 3949 &nd.ni_cnd); 3950 if (error != 0) 3951 goto out; 3952#endif 3953 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3954 NDFREE_PNBUF(&nd); 3955 vput(vp); 3956 if (nd.ni_dvp == vp) 3957 vrele(nd.ni_dvp); 3958 else 3959 vput(nd.ni_dvp); 3960 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH)) != 0) 3961 goto fdout; 3962 goto restart; 3963 } 3964 vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK); 3965 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3966 vn_finished_write(mp); 3967out: 3968 NDFREE_PNBUF(&nd); 3969 vput(vp); 3970 if (nd.ni_dvp == vp) 3971 vrele(nd.ni_dvp); 3972 else 3973 vput(nd.ni_dvp); 3974 if (error == ERELOOKUP) 3975 goto restart; 3976fdout: 3977 if (fp != NULL) 3978 fdrop(fp, td); 3979 return (error); 3980} 3981 3982#if defined(COMPAT_43) || defined(COMPAT_FREEBSD11) 3983int 3984freebsd11_kern_getdirentries(struct thread *td, int fd, char *ubuf, u_int count, 3985 long *basep, void (*func)(struct freebsd11_dirent *)) 3986{ 3987 struct freebsd11_dirent dstdp; 3988 struct dirent *dp, *edp; 3989 char *dirbuf; 3990 off_t base; 3991 ssize_t resid, ucount; 3992 int error; 3993 3994 /* XXX arbitrary sanity limit on `count'. */ 3995 count = min(count, 64 * 1024); 3996 3997 dirbuf = malloc(count, M_TEMP, M_WAITOK); 3998 3999 error = kern_getdirentries(td, fd, dirbuf, count, &base, &resid, 4000 UIO_SYSSPACE); 4001 if (error != 0) 4002 goto done; 4003 if (basep != NULL) 4004 *basep = base; 4005 4006 ucount = 0; 4007 for (dp = (struct dirent *)dirbuf, 4008 edp = (struct dirent *)&dirbuf[count - resid]; 4009 ucount < count && dp < edp; ) { 4010 if (dp->d_reclen == 0) 4011 break; 4012 MPASS(dp->d_reclen >= _GENERIC_DIRLEN(0)); 4013 if (dp->d_namlen >= sizeof(dstdp.d_name)) 4014 continue; 4015 dstdp.d_type = dp->d_type; 4016 dstdp.d_namlen = dp->d_namlen; 4017 dstdp.d_fileno = dp->d_fileno; /* truncate */ 4018 if (dstdp.d_fileno != dp->d_fileno) { 4019 switch (ino64_trunc_error) { 4020 default: 4021 case 0: 4022 break; 4023 case 1: 4024 error = EOVERFLOW; 4025 goto done; 4026 case 2: 4027 dstdp.d_fileno = UINT32_MAX; 4028 break; 4029 } 4030 } 4031 dstdp.d_reclen = sizeof(dstdp) - sizeof(dstdp.d_name) + 4032 ((dp->d_namlen + 1 + 3) &~ 3); 4033 bcopy(dp->d_name, dstdp.d_name, dstdp.d_namlen); 4034 bzero(dstdp.d_name + dstdp.d_namlen, 4035 dstdp.d_reclen - offsetof(struct freebsd11_dirent, d_name) - 4036 dstdp.d_namlen); 4037 MPASS(dstdp.d_reclen <= dp->d_reclen); 4038 MPASS(ucount + dstdp.d_reclen <= count); 4039 if (func != NULL) 4040 func(&dstdp); 4041 error = copyout(&dstdp, ubuf + ucount, dstdp.d_reclen); 4042 if (error != 0) 4043 break; 4044 dp = (struct dirent *)((char *)dp + dp->d_reclen); 4045 ucount += dstdp.d_reclen; 4046 } 4047 4048done: 4049 free(dirbuf, M_TEMP); 4050 if (error == 0) 4051 td->td_retval[0] = ucount; 4052 return (error); 4053} 4054#endif /* COMPAT */ 4055 4056#ifdef COMPAT_43 4057static void 4058ogetdirentries_cvt(struct freebsd11_dirent *dp) 4059{ 4060#if (BYTE_ORDER == LITTLE_ENDIAN) 4061 /* 4062 * The expected low byte of dp->d_namlen is our dp->d_type. 4063 * The high MBZ byte of dp->d_namlen is our dp->d_namlen. 4064 */ 4065 dp->d_type = dp->d_namlen; 4066 dp->d_namlen = 0; 4067#else 4068 /* 4069 * The dp->d_type is the high byte of the expected dp->d_namlen, 4070 * so must be zero'ed. 4071 */ 4072 dp->d_type = 0; 4073#endif 4074} 4075 4076/* 4077 * Read a block of directory entries in a filesystem independent format. 4078 */ 4079#ifndef _SYS_SYSPROTO_H_ 4080struct ogetdirentries_args { 4081 int fd; 4082 char *buf; 4083 u_int count; 4084 long *basep; 4085}; 4086#endif 4087int 4088ogetdirentries(struct thread *td, struct ogetdirentries_args *uap) 4089{ 4090 long loff; 4091 int error; 4092 4093 error = kern_ogetdirentries(td, uap, &loff); 4094 if (error == 0) 4095 error = copyout(&loff, uap->basep, sizeof(long)); 4096 return (error); 4097} 4098 4099int 4100kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap, 4101 long *ploff) 4102{ 4103 long base; 4104 int error; 4105 4106 /* XXX arbitrary sanity limit on `count'. */ 4107 if (uap->count > 64 * 1024) 4108 return (EINVAL); 4109 4110 error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count, 4111 &base, ogetdirentries_cvt); 4112 4113 if (error == 0 && uap->basep != NULL) 4114 error = copyout(&base, uap->basep, sizeof(long)); 4115 4116 return (error); 4117} 4118#endif /* COMPAT_43 */ 4119 4120#if defined(COMPAT_FREEBSD11) 4121#ifndef _SYS_SYSPROTO_H_ 4122struct freebsd11_getdirentries_args { 4123 int fd; 4124 char *buf; 4125 u_int count; 4126 long *basep; 4127}; 4128#endif 4129int 4130freebsd11_getdirentries(struct thread *td, 4131 struct freebsd11_getdirentries_args *uap) 4132{ 4133 long base; 4134 int error; 4135 4136 error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count, 4137 &base, NULL); 4138 4139 if (error == 0 && uap->basep != NULL) 4140 error = copyout(&base, uap->basep, sizeof(long)); 4141 return (error); 4142} 4143 4144int 4145freebsd11_getdents(struct thread *td, struct freebsd11_getdents_args *uap) 4146{ 4147 struct freebsd11_getdirentries_args ap; 4148 4149 ap.fd = uap->fd; 4150 ap.buf = uap->buf; 4151 ap.count = uap->count; 4152 ap.basep = NULL; 4153 return (freebsd11_getdirentries(td, &ap)); 4154} 4155#endif /* COMPAT_FREEBSD11 */ 4156 4157/* 4158 * Read a block of directory entries in a filesystem independent format. 4159 */ 4160int 4161sys_getdirentries(struct thread *td, struct getdirentries_args *uap) 4162{ 4163 off_t base; 4164 int error; 4165 4166 error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base, 4167 NULL, UIO_USERSPACE); 4168 if (error != 0) 4169 return (error); 4170 if (uap->basep != NULL) 4171 error = copyout(&base, uap->basep, sizeof(off_t)); 4172 return (error); 4173} 4174 4175int 4176kern_getdirentries(struct thread *td, int fd, char *buf, size_t count, 4177 off_t *basep, ssize_t *residp, enum uio_seg bufseg) 4178{ 4179 struct vnode *vp; 4180 struct file *fp; 4181 struct uio auio; 4182 struct iovec aiov; 4183 off_t loff; 4184 int error, eofflag; 4185 off_t foffset; 4186 4187 AUDIT_ARG_FD(fd); 4188 if (count > IOSIZE_MAX) 4189 return (EINVAL); 4190 auio.uio_resid = count; 4191 error = getvnode(td, fd, &cap_read_rights, &fp); 4192 if (error != 0) 4193 return (error); 4194 if ((fp->f_flag & FREAD) == 0) { 4195 fdrop(fp, td); 4196 return (EBADF); 4197 } 4198 vp = fp->f_vnode; 4199 foffset = foffset_lock(fp, 0); 4200unionread: 4201 if (vp->v_type != VDIR) { 4202 error = EINVAL; 4203 goto fail; 4204 } 4205 if (__predict_false((vp->v_vflag & VV_UNLINKED) != 0)) { 4206 error = ENOENT; 4207 goto fail; 4208 } 4209 aiov.iov_base = buf; 4210 aiov.iov_len = count; 4211 auio.uio_iov = &aiov; 4212 auio.uio_iovcnt = 1; 4213 auio.uio_rw = UIO_READ; 4214 auio.uio_segflg = bufseg; 4215 auio.uio_td = td; 4216 vn_lock(vp, LK_SHARED | LK_RETRY); 4217 AUDIT_ARG_VNODE1(vp); 4218 loff = auio.uio_offset = foffset; 4219#ifdef MAC 4220 error = mac_vnode_check_readdir(td->td_ucred, vp); 4221 if (error == 0) 4222#endif 4223 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL, 4224 NULL); 4225 foffset = auio.uio_offset; 4226 if (error != 0) { 4227 VOP_UNLOCK(vp); 4228 goto fail; 4229 } 4230 if (count == auio.uio_resid && 4231 (vp->v_vflag & VV_ROOT) && 4232 (vp->v_mount->mnt_flag & MNT_UNION)) { 4233 struct vnode *tvp = vp; 4234 4235 vp = vp->v_mount->mnt_vnodecovered; 4236 VREF(vp); 4237 fp->f_vnode = vp; 4238 foffset = 0; 4239 vput(tvp); 4240 goto unionread; 4241 } 4242 VOP_UNLOCK(vp); 4243 *basep = loff; 4244 if (residp != NULL) 4245 *residp = auio.uio_resid; 4246 td->td_retval[0] = count - auio.uio_resid; 4247fail: 4248 foffset_unlock(fp, foffset, 0); 4249 fdrop(fp, td); 4250 return (error); 4251} 4252 4253/* 4254 * Set the mode mask for creation of filesystem nodes. 4255 */ 4256#ifndef _SYS_SYSPROTO_H_ 4257struct umask_args { 4258 int newmask; 4259}; 4260#endif 4261int 4262sys_umask(struct thread *td, struct umask_args *uap) 4263{ 4264 struct pwddesc *pdp; 4265 4266 pdp = td->td_proc->p_pd; 4267 PWDDESC_XLOCK(pdp); 4268 td->td_retval[0] = pdp->pd_cmask; 4269 pdp->pd_cmask = uap->newmask & ALLPERMS; 4270 PWDDESC_XUNLOCK(pdp); 4271 return (0); 4272} 4273 4274/* 4275 * Void all references to file by ripping underlying filesystem away from 4276 * vnode. 4277 */ 4278#ifndef _SYS_SYSPROTO_H_ 4279struct revoke_args { 4280 char *path; 4281}; 4282#endif 4283int 4284sys_revoke(struct thread *td, struct revoke_args *uap) 4285{ 4286 struct vnode *vp; 4287 struct vattr vattr; 4288 struct nameidata nd; 4289 int error; 4290 4291 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE, 4292 uap->path); 4293 if ((error = namei(&nd)) != 0) 4294 return (error); 4295 vp = nd.ni_vp; 4296 NDFREE_PNBUF(&nd); 4297 if (vp->v_type != VCHR || vp->v_rdev == NULL) { 4298 error = EINVAL; 4299 goto out; 4300 } 4301#ifdef MAC 4302 error = mac_vnode_check_revoke(td->td_ucred, vp); 4303 if (error != 0) 4304 goto out; 4305#endif 4306 error = VOP_GETATTR(vp, &vattr, td->td_ucred); 4307 if (error != 0) 4308 goto out; 4309 if (td->td_ucred->cr_uid != vattr.va_uid) { 4310 error = priv_check(td, PRIV_VFS_ADMIN); 4311 if (error != 0) 4312 goto out; 4313 } 4314 if (devfs_usecount(vp) > 0) 4315 VOP_REVOKE(vp, REVOKEALL); 4316out: 4317 vput(vp); 4318 return (error); 4319} 4320 4321/* 4322 * This variant of getvnode() allows O_PATH files. Caller should 4323 * ensure that returned file and vnode are only used for compatible 4324 * semantics. 4325 */ 4326int 4327getvnode_path(struct thread *td, int fd, cap_rights_t *rightsp, 4328 struct file **fpp) 4329{ 4330 struct file *fp; 4331 int error; 4332 4333 error = fget_unlocked(td, fd, rightsp, &fp); 4334 if (error != 0) 4335 return (error); 4336 4337 /* 4338 * The file could be not of the vnode type, or it may be not 4339 * yet fully initialized, in which case the f_vnode pointer 4340 * may be set, but f_ops is still badfileops. E.g., 4341 * devfs_open() transiently create such situation to 4342 * facilitate csw d_fdopen(). 4343 * 4344 * Dupfdopen() handling in kern_openat() installs the 4345 * half-baked file into the process descriptor table, allowing 4346 * other thread to dereference it. Guard against the race by 4347 * checking f_ops. 4348 */ 4349 if (__predict_false(fp->f_vnode == NULL || fp->f_ops == &badfileops)) { 4350 fdrop(fp, td); 4351 *fpp = NULL; 4352 return (EINVAL); 4353 } 4354 4355 *fpp = fp; 4356 return (0); 4357} 4358 4359/* 4360 * Convert a user file descriptor to a kernel file entry and check 4361 * that, if it is a capability, the correct rights are present. 4362 * A reference on the file entry is held upon returning. 4363 */ 4364int 4365getvnode(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp) 4366{ 4367 int error; 4368 4369 error = getvnode_path(td, fd, rightsp, fpp); 4370 if (__predict_false(error != 0)) 4371 return (error); 4372 4373 /* 4374 * Filter out O_PATH file descriptors, most getvnode() callers 4375 * do not call fo_ methods. 4376 */ 4377 if (__predict_false((*fpp)->f_ops == &path_fileops)) { 4378 fdrop(*fpp, td); 4379 *fpp = NULL; 4380 error = EBADF; 4381 } 4382 4383 return (error); 4384} 4385 4386/* 4387 * Get an (NFS) file handle. 4388 */ 4389#ifndef _SYS_SYSPROTO_H_ 4390struct lgetfh_args { 4391 char *fname; 4392 fhandle_t *fhp; 4393}; 4394#endif 4395int 4396sys_lgetfh(struct thread *td, struct lgetfh_args *uap) 4397{ 4398 4399 return (kern_getfhat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->fname, 4400 UIO_USERSPACE, uap->fhp, UIO_USERSPACE)); 4401} 4402 4403#ifndef _SYS_SYSPROTO_H_ 4404struct getfh_args { 4405 char *fname; 4406 fhandle_t *fhp; 4407}; 4408#endif 4409int 4410sys_getfh(struct thread *td, struct getfh_args *uap) 4411{ 4412 4413 return (kern_getfhat(td, 0, AT_FDCWD, uap->fname, UIO_USERSPACE, 4414 uap->fhp, UIO_USERSPACE)); 4415} 4416 4417/* 4418 * syscall for the rpc.lockd to use to translate an open descriptor into 4419 * a NFS file handle. 4420 * 4421 * warning: do not remove the priv_check() call or this becomes one giant 4422 * security hole. 4423 */ 4424#ifndef _SYS_SYSPROTO_H_ 4425struct getfhat_args { 4426 int fd; 4427 char *path; 4428 fhandle_t *fhp; 4429 int flags; 4430}; 4431#endif 4432int 4433sys_getfhat(struct thread *td, struct getfhat_args *uap) 4434{ 4435 4436 return (kern_getfhat(td, uap->flags, uap->fd, uap->path, UIO_USERSPACE, 4437 uap->fhp, UIO_USERSPACE)); 4438} 4439 4440int 4441kern_getfhat(struct thread *td, int flags, int fd, const char *path, 4442 enum uio_seg pathseg, fhandle_t *fhp, enum uio_seg fhseg) 4443{ 4444 struct nameidata nd; 4445 fhandle_t fh; 4446 struct vnode *vp; 4447 int error; 4448 4449 if ((flags & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH)) != 0) 4450 return (EINVAL); 4451 error = priv_check(td, PRIV_VFS_GETFH); 4452 if (error != 0) 4453 return (error); 4454 NDINIT_AT(&nd, LOOKUP, at2cnpflags(flags, AT_SYMLINK_NOFOLLOW | 4455 AT_RESOLVE_BENEATH) | LOCKLEAF | AUDITVNODE1, pathseg, path, 4456 fd); 4457 error = namei(&nd); 4458 if (error != 0) 4459 return (error); 4460 NDFREE_PNBUF(&nd); 4461 vp = nd.ni_vp; 4462 bzero(&fh, sizeof(fh)); 4463 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid; 4464 error = VOP_VPTOFH(vp, &fh.fh_fid); 4465 vput(vp); 4466 if (error == 0) { 4467 if (fhseg == UIO_USERSPACE) 4468 error = copyout(&fh, fhp, sizeof (fh)); 4469 else 4470 memcpy(fhp, &fh, sizeof(fh)); 4471 } 4472 return (error); 4473} 4474 4475#ifndef _SYS_SYSPROTO_H_ 4476struct fhlink_args { 4477 fhandle_t *fhp; 4478 const char *to; 4479}; 4480#endif 4481int 4482sys_fhlink(struct thread *td, struct fhlink_args *uap) 4483{ 4484 4485 return (kern_fhlinkat(td, AT_FDCWD, uap->to, UIO_USERSPACE, uap->fhp)); 4486} 4487 4488#ifndef _SYS_SYSPROTO_H_ 4489struct fhlinkat_args { 4490 fhandle_t *fhp; 4491 int tofd; 4492 const char *to; 4493}; 4494#endif 4495int 4496sys_fhlinkat(struct thread *td, struct fhlinkat_args *uap) 4497{ 4498 4499 return (kern_fhlinkat(td, uap->tofd, uap->to, UIO_USERSPACE, uap->fhp)); 4500} 4501 4502static int 4503kern_fhlinkat(struct thread *td, int fd, const char *path, 4504 enum uio_seg pathseg, fhandle_t *fhp) 4505{ 4506 fhandle_t fh; 4507 struct mount *mp; 4508 struct vnode *vp; 4509 int error; 4510 4511 error = priv_check(td, PRIV_VFS_GETFH); 4512 if (error != 0) 4513 return (error); 4514 error = copyin(fhp, &fh, sizeof(fh)); 4515 if (error != 0) 4516 return (error); 4517 do { 4518 bwillwrite(); 4519 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL) 4520 return (ESTALE); 4521 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp); 4522 vfs_unbusy(mp); 4523 if (error != 0) 4524 return (error); 4525 VOP_UNLOCK(vp); 4526 error = kern_linkat_vp(td, vp, fd, path, pathseg); 4527 } while (error == EAGAIN || error == ERELOOKUP); 4528 return (error); 4529} 4530 4531#ifndef _SYS_SYSPROTO_H_ 4532struct fhreadlink_args { 4533 fhandle_t *fhp; 4534 char *buf; 4535 size_t bufsize; 4536}; 4537#endif 4538int 4539sys_fhreadlink(struct thread *td, struct fhreadlink_args *uap) 4540{ 4541 fhandle_t fh; 4542 struct mount *mp; 4543 struct vnode *vp; 4544 int error; 4545 4546 error = priv_check(td, PRIV_VFS_GETFH); 4547 if (error != 0) 4548 return (error); 4549 if (uap->bufsize > IOSIZE_MAX) 4550 return (EINVAL); 4551 error = copyin(uap->fhp, &fh, sizeof(fh)); 4552 if (error != 0) 4553 return (error); 4554 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL) 4555 return (ESTALE); 4556 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp); 4557 vfs_unbusy(mp); 4558 if (error != 0) 4559 return (error); 4560 error = kern_readlink_vp(vp, uap->buf, UIO_USERSPACE, uap->bufsize, td); 4561 vput(vp); 4562 return (error); 4563} 4564 4565/* 4566 * syscall for the rpc.lockd to use to translate a NFS file handle into an 4567 * open descriptor. 4568 * 4569 * warning: do not remove the priv_check() call or this becomes one giant 4570 * security hole. 4571 */ 4572#ifndef _SYS_SYSPROTO_H_ 4573struct fhopen_args { 4574 const struct fhandle *u_fhp; 4575 int flags; 4576}; 4577#endif 4578int 4579sys_fhopen(struct thread *td, struct fhopen_args *uap) 4580{ 4581 return (kern_fhopen(td, uap->u_fhp, uap->flags)); 4582} 4583 4584int 4585kern_fhopen(struct thread *td, const struct fhandle *u_fhp, int flags) 4586{ 4587 struct mount *mp; 4588 struct vnode *vp; 4589 struct fhandle fhp; 4590 struct file *fp; 4591 int fmode, error; 4592 int indx; 4593 4594 error = priv_check(td, PRIV_VFS_FHOPEN); 4595 if (error != 0) 4596 return (error); 4597 indx = -1; 4598 fmode = FFLAGS(flags); 4599 /* why not allow a non-read/write open for our lockd? */ 4600 if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT)) 4601 return (EINVAL); 4602 error = copyin(u_fhp, &fhp, sizeof(fhp)); 4603 if (error != 0) 4604 return(error); 4605 /* find the mount point */ 4606 mp = vfs_busyfs(&fhp.fh_fsid); 4607 if (mp == NULL) 4608 return (ESTALE); 4609 /* now give me my vnode, it gets returned to me locked */ 4610 error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp); 4611 vfs_unbusy(mp); 4612 if (error != 0) 4613 return (error); 4614 4615 error = falloc_noinstall(td, &fp); 4616 if (error != 0) { 4617 vput(vp); 4618 return (error); 4619 } 4620 /* 4621 * An extra reference on `fp' has been held for us by 4622 * falloc_noinstall(). 4623 */ 4624 4625#ifdef INVARIANTS 4626 td->td_dupfd = -1; 4627#endif 4628 error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp); 4629 if (error != 0) { 4630 KASSERT(fp->f_ops == &badfileops, 4631 ("VOP_OPEN in fhopen() set f_ops")); 4632 KASSERT(td->td_dupfd < 0, 4633 ("fhopen() encountered fdopen()")); 4634 4635 vput(vp); 4636 goto bad; 4637 } 4638#ifdef INVARIANTS 4639 td->td_dupfd = 0; 4640#endif 4641 fp->f_vnode = vp; 4642 finit_vnode(fp, fmode, NULL, &vnops); 4643 VOP_UNLOCK(vp); 4644 if ((fmode & O_TRUNC) != 0) { 4645 error = fo_truncate(fp, 0, td->td_ucred, td); 4646 if (error != 0) 4647 goto bad; 4648 } 4649 4650 error = finstall(td, fp, &indx, fmode, NULL); 4651bad: 4652 fdrop(fp, td); 4653 td->td_retval[0] = indx; 4654 return (error); 4655} 4656 4657/* 4658 * Stat an (NFS) file handle. 4659 */ 4660#ifndef _SYS_SYSPROTO_H_ 4661struct fhstat_args { 4662 struct fhandle *u_fhp; 4663 struct stat *sb; 4664}; 4665#endif 4666int 4667sys_fhstat(struct thread *td, struct fhstat_args *uap) 4668{ 4669 struct stat sb; 4670 struct fhandle fh; 4671 int error; 4672 4673 error = copyin(uap->u_fhp, &fh, sizeof(fh)); 4674 if (error != 0) 4675 return (error); 4676 error = kern_fhstat(td, fh, &sb); 4677 if (error == 0) 4678 error = copyout(&sb, uap->sb, sizeof(sb)); 4679 return (error); 4680} 4681 4682int 4683kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb) 4684{ 4685 struct mount *mp; 4686 struct vnode *vp; 4687 int error; 4688 4689 error = priv_check(td, PRIV_VFS_FHSTAT); 4690 if (error != 0) 4691 return (error); 4692 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL) 4693 return (ESTALE); 4694 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp); 4695 vfs_unbusy(mp); 4696 if (error != 0) 4697 return (error); 4698 error = VOP_STAT(vp, sb, td->td_ucred, NOCRED); 4699 vput(vp); 4700 return (error); 4701} 4702 4703/* 4704 * Implement fstatfs() for (NFS) file handles. 4705 */ 4706#ifndef _SYS_SYSPROTO_H_ 4707struct fhstatfs_args { 4708 struct fhandle *u_fhp; 4709 struct statfs *buf; 4710}; 4711#endif 4712int 4713sys_fhstatfs(struct thread *td, struct fhstatfs_args *uap) 4714{ 4715 struct statfs *sfp; 4716 fhandle_t fh; 4717 int error; 4718 4719 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t)); 4720 if (error != 0) 4721 return (error); 4722 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 4723 error = kern_fhstatfs(td, fh, sfp); 4724 if (error == 0) 4725 error = copyout(sfp, uap->buf, sizeof(*sfp)); 4726 free(sfp, M_STATFS); 4727 return (error); 4728} 4729 4730int 4731kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf) 4732{ 4733 struct mount *mp; 4734 struct vnode *vp; 4735 int error; 4736 4737 error = priv_check(td, PRIV_VFS_FHSTATFS); 4738 if (error != 0) 4739 return (error); 4740 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL) 4741 return (ESTALE); 4742 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp); 4743 if (error != 0) { 4744 vfs_unbusy(mp); 4745 return (error); 4746 } 4747 vput(vp); 4748 error = prison_canseemount(td->td_ucred, mp); 4749 if (error != 0) 4750 goto out; 4751#ifdef MAC 4752 error = mac_mount_check_stat(td->td_ucred, mp); 4753 if (error != 0) 4754 goto out; 4755#endif 4756 error = VFS_STATFS(mp, buf); 4757out: 4758 vfs_unbusy(mp); 4759 return (error); 4760} 4761 4762/* 4763 * Unlike madvise(2), we do not make a best effort to remember every 4764 * possible caching hint. Instead, we remember the last setting with 4765 * the exception that we will allow POSIX_FADV_NORMAL to adjust the 4766 * region of any current setting. 4767 */ 4768int 4769kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len, 4770 int advice) 4771{ 4772 struct fadvise_info *fa, *new; 4773 struct file *fp; 4774 struct vnode *vp; 4775 off_t end; 4776 int error; 4777 4778 if (offset < 0 || len < 0 || offset > OFF_MAX - len) 4779 return (EINVAL); 4780 AUDIT_ARG_VALUE(advice); 4781 switch (advice) { 4782 case POSIX_FADV_SEQUENTIAL: 4783 case POSIX_FADV_RANDOM: 4784 case POSIX_FADV_NOREUSE: 4785 new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK); 4786 break; 4787 case POSIX_FADV_NORMAL: 4788 case POSIX_FADV_WILLNEED: 4789 case POSIX_FADV_DONTNEED: 4790 new = NULL; 4791 break; 4792 default: 4793 return (EINVAL); 4794 } 4795 /* XXX: CAP_POSIX_FADVISE? */ 4796 AUDIT_ARG_FD(fd); 4797 error = fget(td, fd, &cap_no_rights, &fp); 4798 if (error != 0) 4799 goto out; 4800 AUDIT_ARG_FILE(td->td_proc, fp); 4801 if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) { 4802 error = ESPIPE; 4803 goto out; 4804 } 4805 if (fp->f_type != DTYPE_VNODE) { 4806 error = ENODEV; 4807 goto out; 4808 } 4809 vp = fp->f_vnode; 4810 if (vp->v_type != VREG) { 4811 error = ENODEV; 4812 goto out; 4813 } 4814 if (len == 0) 4815 end = OFF_MAX; 4816 else 4817 end = offset + len - 1; 4818 switch (advice) { 4819 case POSIX_FADV_SEQUENTIAL: 4820 case POSIX_FADV_RANDOM: 4821 case POSIX_FADV_NOREUSE: 4822 /* 4823 * Try to merge any existing non-standard region with 4824 * this new region if possible, otherwise create a new 4825 * non-standard region for this request. 4826 */ 4827 mtx_pool_lock(mtxpool_sleep, fp); 4828 fa = fp->f_advice; 4829 if (fa != NULL && fa->fa_advice == advice && 4830 ((fa->fa_start <= end && fa->fa_end >= offset) || 4831 (end != OFF_MAX && fa->fa_start == end + 1) || 4832 (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) { 4833 if (offset < fa->fa_start) 4834 fa->fa_start = offset; 4835 if (end > fa->fa_end) 4836 fa->fa_end = end; 4837 } else { 4838 new->fa_advice = advice; 4839 new->fa_start = offset; 4840 new->fa_end = end; 4841 fp->f_advice = new; 4842 new = fa; 4843 } 4844 mtx_pool_unlock(mtxpool_sleep, fp); 4845 break; 4846 case POSIX_FADV_NORMAL: 4847 /* 4848 * If a the "normal" region overlaps with an existing 4849 * non-standard region, trim or remove the 4850 * non-standard region. 4851 */ 4852 mtx_pool_lock(mtxpool_sleep, fp); 4853 fa = fp->f_advice; 4854 if (fa != NULL) { 4855 if (offset <= fa->fa_start && end >= fa->fa_end) { 4856 new = fa; 4857 fp->f_advice = NULL; 4858 } else if (offset <= fa->fa_start && 4859 end >= fa->fa_start) 4860 fa->fa_start = end + 1; 4861 else if (offset <= fa->fa_end && end >= fa->fa_end) 4862 fa->fa_end = offset - 1; 4863 else if (offset >= fa->fa_start && end <= fa->fa_end) { 4864 /* 4865 * If the "normal" region is a middle 4866 * portion of the existing 4867 * non-standard region, just remove 4868 * the whole thing rather than picking 4869 * one side or the other to 4870 * preserve. 4871 */ 4872 new = fa; 4873 fp->f_advice = NULL; 4874 } 4875 } 4876 mtx_pool_unlock(mtxpool_sleep, fp); 4877 break; 4878 case POSIX_FADV_WILLNEED: 4879 case POSIX_FADV_DONTNEED: 4880 error = VOP_ADVISE(vp, offset, end, advice); 4881 break; 4882 } 4883out: 4884 if (fp != NULL) 4885 fdrop(fp, td); 4886 free(new, M_FADVISE); 4887 return (error); 4888} 4889 4890int 4891sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap) 4892{ 4893 int error; 4894 4895 error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len, 4896 uap->advice); 4897 return (kern_posix_error(td, error)); 4898} 4899 4900int 4901kern_copy_file_range(struct thread *td, int infd, off_t *inoffp, int outfd, 4902 off_t *outoffp, size_t len, unsigned int flags) 4903{ 4904 struct file *infp, *outfp; 4905 struct vnode *invp, *outvp; 4906 int error; 4907 size_t retlen; 4908 void *rl_rcookie, *rl_wcookie; 4909 off_t savinoff, savoutoff; 4910 4911 infp = outfp = NULL; 4912 rl_rcookie = rl_wcookie = NULL; 4913 savinoff = -1; 4914 error = 0; 4915 retlen = 0; 4916 4917 if (flags != 0) { 4918 error = EINVAL; 4919 goto out; 4920 } 4921 if (len > SSIZE_MAX) 4922 /* 4923 * Although the len argument is size_t, the return argument 4924 * is ssize_t (which is signed). Therefore a size that won't 4925 * fit in ssize_t can't be returned. 4926 */ 4927 len = SSIZE_MAX; 4928 4929 /* Get the file structures for the file descriptors. */ 4930 error = fget_read(td, infd, 4931 inoffp != NULL ? &cap_pread_rights : &cap_read_rights, &infp); 4932 if (error != 0) 4933 goto out; 4934 if (infp->f_ops == &badfileops) { 4935 error = EBADF; 4936 goto out; 4937 } 4938 if (infp->f_vnode == NULL) { 4939 error = EINVAL; 4940 goto out; 4941 } 4942 error = fget_write(td, outfd, 4943 outoffp != NULL ? &cap_pwrite_rights : &cap_write_rights, &outfp); 4944 if (error != 0) 4945 goto out; 4946 if (outfp->f_ops == &badfileops) { 4947 error = EBADF; 4948 goto out; 4949 } 4950 if (outfp->f_vnode == NULL) { 4951 error = EINVAL; 4952 goto out; 4953 } 4954 4955 /* Set the offset pointers to the correct place. */ 4956 if (inoffp == NULL) 4957 inoffp = &infp->f_offset; 4958 if (outoffp == NULL) 4959 outoffp = &outfp->f_offset; 4960 savinoff = *inoffp; 4961 savoutoff = *outoffp; 4962 4963 invp = infp->f_vnode; 4964 outvp = outfp->f_vnode; 4965 /* Sanity check the f_flag bits. */ 4966 if ((outfp->f_flag & (FWRITE | FAPPEND)) != FWRITE || 4967 (infp->f_flag & FREAD) == 0) { 4968 error = EBADF; 4969 goto out; 4970 } 4971 4972 /* If len == 0, just return 0. */ 4973 if (len == 0) 4974 goto out; 4975 4976 /* 4977 * If infp and outfp refer to the same file, the byte ranges cannot 4978 * overlap. 4979 */ 4980 if (invp == outvp && ((savinoff <= savoutoff && savinoff + len > 4981 savoutoff) || (savinoff > savoutoff && savoutoff + len > 4982 savinoff))) { 4983 error = EINVAL; 4984 goto out; 4985 } 4986 4987 /* Range lock the byte ranges for both invp and outvp. */ 4988 for (;;) { 4989 rl_wcookie = vn_rangelock_wlock(outvp, *outoffp, *outoffp + 4990 len); 4991 rl_rcookie = vn_rangelock_tryrlock(invp, *inoffp, *inoffp + 4992 len); 4993 if (rl_rcookie != NULL) 4994 break; 4995 vn_rangelock_unlock(outvp, rl_wcookie); 4996 rl_rcookie = vn_rangelock_rlock(invp, *inoffp, *inoffp + len); 4997 vn_rangelock_unlock(invp, rl_rcookie); 4998 } 4999 5000 retlen = len; 5001 error = vn_copy_file_range(invp, inoffp, outvp, outoffp, &retlen, 5002 flags, infp->f_cred, outfp->f_cred, td); 5003out: 5004 if (rl_rcookie != NULL) 5005 vn_rangelock_unlock(invp, rl_rcookie); 5006 if (rl_wcookie != NULL) 5007 vn_rangelock_unlock(outvp, rl_wcookie); 5008 if (savinoff != -1 && (error == EINTR || error == ERESTART)) { 5009 *inoffp = savinoff; 5010 *outoffp = savoutoff; 5011 } 5012 if (outfp != NULL) 5013 fdrop(outfp, td); 5014 if (infp != NULL) 5015 fdrop(infp, td); 5016 td->td_retval[0] = retlen; 5017 return (error); 5018} 5019 5020int 5021sys_copy_file_range(struct thread *td, struct copy_file_range_args *uap) 5022{ 5023 off_t inoff, outoff, *inoffp, *outoffp; 5024 int error; 5025 5026 inoffp = outoffp = NULL; 5027 if (uap->inoffp != NULL) { 5028 error = copyin(uap->inoffp, &inoff, sizeof(off_t)); 5029 if (error != 0) 5030 return (error); 5031 inoffp = &inoff; 5032 } 5033 if (uap->outoffp != NULL) { 5034 error = copyin(uap->outoffp, &outoff, sizeof(off_t)); 5035 if (error != 0) 5036 return (error); 5037 outoffp = &outoff; 5038 } 5039 error = kern_copy_file_range(td, uap->infd, inoffp, uap->outfd, 5040 outoffp, uap->len, uap->flags); 5041 if (error == 0 && uap->inoffp != NULL) 5042 error = copyout(inoffp, uap->inoffp, sizeof(off_t)); 5043 if (error == 0 && uap->outoffp != NULL) 5044 error = copyout(outoffp, uap->outoffp, sizeof(off_t)); 5045 return (error); 5046} 5047