1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1989, 1993 5 * The Regents of the University of California. All rights reserved. 6 * (c) UNIX System Laboratories, Inc. 7 * All or some portions of this file are derived from material licensed 8 * to the University of California by American Telephone and Telegraph 9 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 10 * the permission of UNIX System Laboratories, Inc. 11 * 12 * Redistribution and use in source and binary forms, with or without 13 * modification, are permitted provided that the following conditions 14 * are met: 15 * 1. Redistributions of source code must retain the above copyright 16 * notice, this list of conditions and the following disclaimer. 17 * 2. Redistributions in binary form must reproduce the above copyright 18 * notice, this list of conditions and the following disclaimer in the 19 * documentation and/or other materials provided with the distribution. 20 * 3. Neither the name of the University nor the names of its contributors 21 * may be used to endorse or promote products derived from this software 22 * without specific prior written permission. 23 * 24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 34 * SUCH DAMAGE. 35 * 36 * @(#)vfs_syscalls.c 8.13 (Berkeley) 4/15/94 37 */ 38 39 #include <sys/cdefs.h> 40 __FBSDID("$FreeBSD$"); 41 42 #include "opt_capsicum.h" 43 #include "opt_ktrace.h" 44 45 #include <sys/param.h> 46 #include <sys/systm.h> 47 #include <sys/bio.h> 48 #include <sys/buf.h> 49 #include <sys/capsicum.h> 50 #include <sys/disk.h> 51 #include <sys/sysent.h> 52 #include <sys/malloc.h> 53 #include <sys/mount.h> 54 #include <sys/mutex.h> 55 #include <sys/sysproto.h> 56 #include <sys/namei.h> 57 #include <sys/filedesc.h> 58 #include <sys/kernel.h> 59 #include <sys/fcntl.h> 60 #include <sys/file.h> 61 #include <sys/filio.h> 62 #include <sys/limits.h> 63 #include <sys/linker.h> 64 #include <sys/rwlock.h> 65 #include <sys/sdt.h> 66 #include <sys/stat.h> 67 #include <sys/sx.h> 68 #include <sys/unistd.h> 69 #include <sys/vnode.h> 70 #include <sys/priv.h> 71 #include <sys/proc.h> 72 #include <sys/dirent.h> 73 #include <sys/jail.h> 74 #include <sys/syscallsubr.h> 75 #include <sys/sysctl.h> 76 #ifdef KTRACE 77 #include <sys/ktrace.h> 78 #endif 79 80 #include <machine/stdarg.h> 81 82 #include <security/audit/audit.h> 83 #include <security/mac/mac_framework.h> 84 85 #include <vm/vm.h> 86 #include <vm/vm_object.h> 87 #include <vm/vm_page.h> 88 #include <vm/uma.h> 89 90 #include <ufs/ufs/quota.h> 91 92 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information"); 93 94 SDT_PROVIDER_DEFINE(vfs); 95 SDT_PROBE_DEFINE2(vfs, , stat, mode, "char *", "int"); 96 SDT_PROBE_DEFINE2(vfs, , stat, reg, "char *", "int"); 97 98 static int kern_chflagsat(struct thread *td, int fd, const char *path, 99 enum uio_seg pathseg, u_long flags, int atflag); 100 static int setfflags(struct thread *td, struct vnode *, u_long); 101 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *); 102 static int getutimens(const struct timespec *, enum uio_seg, 103 struct timespec *, int *); 104 static int setutimes(struct thread *td, struct vnode *, 105 const struct timespec *, int, int); 106 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred, 107 struct thread *td); 108 static int kern_fhlinkat(struct thread *td, int fd, const char *path, 109 enum uio_seg pathseg, fhandle_t *fhp); 110 static int kern_getfhat(struct thread *td, int flags, int fd, 111 const char *path, enum uio_seg pathseg, fhandle_t *fhp); 112 static int kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg, 113 size_t count, struct thread *td); 114 static int kern_linkat_vp(struct thread *td, struct vnode *vp, int fd, 115 const char *path, enum uio_seg segflag); 116 117 int 118 kern_sync(struct thread *td) 119 { 120 struct mount *mp, *nmp; 121 int save; 122 123 mtx_lock(&mountlist_mtx); 124 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) { 125 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) { 126 nmp = TAILQ_NEXT(mp, mnt_list); 127 continue; 128 } 129 if ((mp->mnt_flag & MNT_RDONLY) == 0 && 130 vn_start_write(NULL, &mp, V_NOWAIT) == 0) { 131 save = curthread_pflags_set(TDP_SYNCIO); 132 vfs_periodic(mp, MNT_NOWAIT); 133 VFS_SYNC(mp, MNT_NOWAIT); 134 curthread_pflags_restore(save); 135 vn_finished_write(mp); 136 } 137 mtx_lock(&mountlist_mtx); 138 nmp = TAILQ_NEXT(mp, mnt_list); 139 vfs_unbusy(mp); 140 } 141 mtx_unlock(&mountlist_mtx); 142 return (0); 143 } 144 145 /* 146 * Sync each mounted filesystem. 147 */ 148 #ifndef _SYS_SYSPROTO_H_ 149 struct sync_args { 150 int dummy; 151 }; 152 #endif 153 /* ARGSUSED */ 154 int 155 sys_sync(struct thread *td, struct sync_args *uap) 156 { 157 158 return (kern_sync(td)); 159 } 160 161 /* 162 * Change filesystem quotas. 163 */ 164 #ifndef _SYS_SYSPROTO_H_ 165 struct quotactl_args { 166 char *path; 167 int cmd; 168 int uid; 169 caddr_t arg; 170 }; 171 #endif 172 int 173 sys_quotactl(struct thread *td, struct quotactl_args *uap) 174 { 175 struct mount *mp; 176 struct nameidata nd; 177 int error; 178 179 AUDIT_ARG_CMD(uap->cmd); 180 AUDIT_ARG_UID(uap->uid); 181 if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS)) 182 return (EPERM); 183 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE, 184 uap->path, td); 185 if ((error = namei(&nd)) != 0) 186 return (error); 187 NDFREE(&nd, NDF_ONLY_PNBUF); 188 mp = nd.ni_vp->v_mount; 189 vfs_ref(mp); 190 vput(nd.ni_vp); 191 error = vfs_busy(mp, 0); 192 if (error != 0) { 193 vfs_rel(mp); 194 return (error); 195 } 196 error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg); 197 198 /* 199 * Since quota on operation typically needs to open quota 200 * file, the Q_QUOTAON handler needs to unbusy the mount point 201 * before calling into namei. Otherwise, unmount might be 202 * started between two vfs_busy() invocations (first is our, 203 * second is from mount point cross-walk code in lookup()), 204 * causing deadlock. 205 * 206 * Require that Q_QUOTAON handles the vfs_busy() reference on 207 * its own, always returning with ubusied mount point. 208 */ 209 if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON && 210 (uap->cmd >> SUBCMDSHIFT) != Q_QUOTAOFF) 211 vfs_unbusy(mp); 212 vfs_rel(mp); 213 return (error); 214 } 215 216 /* 217 * Used by statfs conversion routines to scale the block size up if 218 * necessary so that all of the block counts are <= 'max_size'. Note 219 * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero 220 * value of 'n'. 221 */ 222 void 223 statfs_scale_blocks(struct statfs *sf, long max_size) 224 { 225 uint64_t count; 226 int shift; 227 228 KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__)); 229 230 /* 231 * Attempt to scale the block counts to give a more accurate 232 * overview to userland of the ratio of free space to used 233 * space. To do this, find the largest block count and compute 234 * a divisor that lets it fit into a signed integer <= max_size. 235 */ 236 if (sf->f_bavail < 0) 237 count = -sf->f_bavail; 238 else 239 count = sf->f_bavail; 240 count = MAX(sf->f_blocks, MAX(sf->f_bfree, count)); 241 if (count <= max_size) 242 return; 243 244 count >>= flsl(max_size); 245 shift = 0; 246 while (count > 0) { 247 shift++; 248 count >>=1; 249 } 250 251 sf->f_bsize <<= shift; 252 sf->f_blocks >>= shift; 253 sf->f_bfree >>= shift; 254 sf->f_bavail >>= shift; 255 } 256 257 static int 258 kern_do_statfs(struct thread *td, struct mount *mp, struct statfs *buf) 259 { 260 int error; 261 262 if (mp == NULL) 263 return (EBADF); 264 error = vfs_busy(mp, 0); 265 vfs_rel(mp); 266 if (error != 0) 267 return (error); 268 #ifdef MAC 269 error = mac_mount_check_stat(td->td_ucred, mp); 270 if (error != 0) 271 goto out; 272 #endif 273 error = VFS_STATFS(mp, buf); 274 if (error != 0) 275 goto out; 276 if (priv_check_cred_vfs_generation(td->td_ucred)) { 277 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0; 278 prison_enforce_statfs(td->td_ucred, mp, buf); 279 } 280 out: 281 vfs_unbusy(mp); 282 return (error); 283 } 284 285 /* 286 * Get filesystem statistics. 287 */ 288 #ifndef _SYS_SYSPROTO_H_ 289 struct statfs_args { 290 char *path; 291 struct statfs *buf; 292 }; 293 #endif 294 int 295 sys_statfs(struct thread *td, struct statfs_args *uap) 296 { 297 struct statfs *sfp; 298 int error; 299 300 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 301 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp); 302 if (error == 0) 303 error = copyout(sfp, uap->buf, sizeof(struct statfs)); 304 free(sfp, M_STATFS); 305 return (error); 306 } 307 308 int 309 kern_statfs(struct thread *td, const char *path, enum uio_seg pathseg, 310 struct statfs *buf) 311 { 312 struct mount *mp; 313 struct nameidata nd; 314 int error; 315 316 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1, 317 pathseg, path, td); 318 error = namei(&nd); 319 if (error != 0) 320 return (error); 321 mp = nd.ni_vp->v_mount; 322 vfs_ref(mp); 323 NDFREE(&nd, NDF_ONLY_PNBUF); 324 vput(nd.ni_vp); 325 return (kern_do_statfs(td, mp, buf)); 326 } 327 328 /* 329 * Get filesystem statistics. 330 */ 331 #ifndef _SYS_SYSPROTO_H_ 332 struct fstatfs_args { 333 int fd; 334 struct statfs *buf; 335 }; 336 #endif 337 int 338 sys_fstatfs(struct thread *td, struct fstatfs_args *uap) 339 { 340 struct statfs *sfp; 341 int error; 342 343 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 344 error = kern_fstatfs(td, uap->fd, sfp); 345 if (error == 0) 346 error = copyout(sfp, uap->buf, sizeof(struct statfs)); 347 free(sfp, M_STATFS); 348 return (error); 349 } 350 351 int 352 kern_fstatfs(struct thread *td, int fd, struct statfs *buf) 353 { 354 struct file *fp; 355 struct mount *mp; 356 struct vnode *vp; 357 int error; 358 359 AUDIT_ARG_FD(fd); 360 error = getvnode(td, fd, &cap_fstatfs_rights, &fp); 361 if (error != 0) 362 return (error); 363 vp = fp->f_vnode; 364 vn_lock(vp, LK_SHARED | LK_RETRY); 365 #ifdef AUDIT 366 AUDIT_ARG_VNODE1(vp); 367 #endif 368 mp = vp->v_mount; 369 if (mp != NULL) 370 vfs_ref(mp); 371 VOP_UNLOCK(vp); 372 fdrop(fp, td); 373 return (kern_do_statfs(td, mp, buf)); 374 } 375 376 /* 377 * Get statistics on all filesystems. 378 */ 379 #ifndef _SYS_SYSPROTO_H_ 380 struct getfsstat_args { 381 struct statfs *buf; 382 long bufsize; 383 int mode; 384 }; 385 #endif 386 int 387 sys_getfsstat(struct thread *td, struct getfsstat_args *uap) 388 { 389 size_t count; 390 int error; 391 392 if (uap->bufsize < 0 || uap->bufsize > SIZE_MAX) 393 return (EINVAL); 394 error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count, 395 UIO_USERSPACE, uap->mode); 396 if (error == 0) 397 td->td_retval[0] = count; 398 return (error); 399 } 400 401 /* 402 * If (bufsize > 0 && bufseg == UIO_SYSSPACE) 403 * The caller is responsible for freeing memory which will be allocated 404 * in '*buf'. 405 */ 406 int 407 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize, 408 size_t *countp, enum uio_seg bufseg, int mode) 409 { 410 struct mount *mp, *nmp; 411 struct statfs *sfsp, *sp, *sptmp, *tofree; 412 size_t count, maxcount; 413 int error; 414 415 switch (mode) { 416 case MNT_WAIT: 417 case MNT_NOWAIT: 418 break; 419 default: 420 if (bufseg == UIO_SYSSPACE) 421 *buf = NULL; 422 return (EINVAL); 423 } 424 restart: 425 maxcount = bufsize / sizeof(struct statfs); 426 if (bufsize == 0) { 427 sfsp = NULL; 428 tofree = NULL; 429 } else if (bufseg == UIO_USERSPACE) { 430 sfsp = *buf; 431 tofree = NULL; 432 } else /* if (bufseg == UIO_SYSSPACE) */ { 433 count = 0; 434 mtx_lock(&mountlist_mtx); 435 TAILQ_FOREACH(mp, &mountlist, mnt_list) { 436 count++; 437 } 438 mtx_unlock(&mountlist_mtx); 439 if (maxcount > count) 440 maxcount = count; 441 tofree = sfsp = *buf = malloc(maxcount * sizeof(struct statfs), 442 M_STATFS, M_WAITOK); 443 } 444 445 count = 0; 446 447 /* 448 * If there is no target buffer they only want the count. 449 * 450 * This could be TAILQ_FOREACH but it is open-coded to match the original 451 * code below. 452 */ 453 if (sfsp == NULL) { 454 mtx_lock(&mountlist_mtx); 455 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) { 456 if (prison_canseemount(td->td_ucred, mp) != 0) { 457 nmp = TAILQ_NEXT(mp, mnt_list); 458 continue; 459 } 460 #ifdef MAC 461 if (mac_mount_check_stat(td->td_ucred, mp) != 0) { 462 nmp = TAILQ_NEXT(mp, mnt_list); 463 continue; 464 } 465 #endif 466 count++; 467 nmp = TAILQ_NEXT(mp, mnt_list); 468 } 469 mtx_unlock(&mountlist_mtx); 470 *countp = count; 471 return (0); 472 } 473 474 /* 475 * They want the entire thing. 476 * 477 * Short-circuit the corner case of no room for anything, avoids 478 * relocking below. 479 */ 480 if (maxcount < 1) { 481 goto out; 482 } 483 484 mtx_lock(&mountlist_mtx); 485 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) { 486 if (prison_canseemount(td->td_ucred, mp) != 0) { 487 nmp = TAILQ_NEXT(mp, mnt_list); 488 continue; 489 } 490 #ifdef MAC 491 if (mac_mount_check_stat(td->td_ucred, mp) != 0) { 492 nmp = TAILQ_NEXT(mp, mnt_list); 493 continue; 494 } 495 #endif 496 if (mode == MNT_WAIT) { 497 if (vfs_busy(mp, MBF_MNTLSTLOCK) != 0) { 498 /* 499 * If vfs_busy() failed, and MBF_NOWAIT 500 * wasn't passed, then the mp is gone. 501 * Furthermore, because of MBF_MNTLSTLOCK, 502 * the mountlist_mtx was dropped. We have 503 * no other choice than to start over. 504 */ 505 mtx_unlock(&mountlist_mtx); 506 free(tofree, M_STATFS); 507 goto restart; 508 } 509 } else { 510 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK) != 0) { 511 nmp = TAILQ_NEXT(mp, mnt_list); 512 continue; 513 } 514 } 515 sp = &mp->mnt_stat; 516 /* 517 * If MNT_NOWAIT is specified, do not refresh 518 * the fsstat cache. 519 */ 520 if (mode != MNT_NOWAIT) { 521 error = VFS_STATFS(mp, sp); 522 if (error != 0) { 523 mtx_lock(&mountlist_mtx); 524 nmp = TAILQ_NEXT(mp, mnt_list); 525 vfs_unbusy(mp); 526 continue; 527 } 528 } 529 if (priv_check_cred_vfs_generation(td->td_ucred)) { 530 sptmp = malloc(sizeof(struct statfs), M_STATFS, 531 M_WAITOK); 532 *sptmp = *sp; 533 sptmp->f_fsid.val[0] = sptmp->f_fsid.val[1] = 0; 534 prison_enforce_statfs(td->td_ucred, mp, sptmp); 535 sp = sptmp; 536 } else 537 sptmp = NULL; 538 if (bufseg == UIO_SYSSPACE) { 539 bcopy(sp, sfsp, sizeof(*sp)); 540 free(sptmp, M_STATFS); 541 } else /* if (bufseg == UIO_USERSPACE) */ { 542 error = copyout(sp, sfsp, sizeof(*sp)); 543 free(sptmp, M_STATFS); 544 if (error != 0) { 545 vfs_unbusy(mp); 546 return (error); 547 } 548 } 549 sfsp++; 550 count++; 551 552 if (count == maxcount) { 553 vfs_unbusy(mp); 554 goto out; 555 } 556 557 mtx_lock(&mountlist_mtx); 558 nmp = TAILQ_NEXT(mp, mnt_list); 559 vfs_unbusy(mp); 560 } 561 mtx_unlock(&mountlist_mtx); 562 out: 563 *countp = count; 564 return (0); 565 } 566 567 #ifdef COMPAT_FREEBSD4 568 /* 569 * Get old format filesystem statistics. 570 */ 571 static void freebsd4_cvtstatfs(struct statfs *, struct ostatfs *); 572 573 #ifndef _SYS_SYSPROTO_H_ 574 struct freebsd4_statfs_args { 575 char *path; 576 struct ostatfs *buf; 577 }; 578 #endif 579 int 580 freebsd4_statfs(struct thread *td, struct freebsd4_statfs_args *uap) 581 { 582 struct ostatfs osb; 583 struct statfs *sfp; 584 int error; 585 586 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 587 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp); 588 if (error == 0) { 589 freebsd4_cvtstatfs(sfp, &osb); 590 error = copyout(&osb, uap->buf, sizeof(osb)); 591 } 592 free(sfp, M_STATFS); 593 return (error); 594 } 595 596 /* 597 * Get filesystem statistics. 598 */ 599 #ifndef _SYS_SYSPROTO_H_ 600 struct freebsd4_fstatfs_args { 601 int fd; 602 struct ostatfs *buf; 603 }; 604 #endif 605 int 606 freebsd4_fstatfs(struct thread *td, struct freebsd4_fstatfs_args *uap) 607 { 608 struct ostatfs osb; 609 struct statfs *sfp; 610 int error; 611 612 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 613 error = kern_fstatfs(td, uap->fd, sfp); 614 if (error == 0) { 615 freebsd4_cvtstatfs(sfp, &osb); 616 error = copyout(&osb, uap->buf, sizeof(osb)); 617 } 618 free(sfp, M_STATFS); 619 return (error); 620 } 621 622 /* 623 * Get statistics on all filesystems. 624 */ 625 #ifndef _SYS_SYSPROTO_H_ 626 struct freebsd4_getfsstat_args { 627 struct ostatfs *buf; 628 long bufsize; 629 int mode; 630 }; 631 #endif 632 int 633 freebsd4_getfsstat(struct thread *td, struct freebsd4_getfsstat_args *uap) 634 { 635 struct statfs *buf, *sp; 636 struct ostatfs osb; 637 size_t count, size; 638 int error; 639 640 if (uap->bufsize < 0) 641 return (EINVAL); 642 count = uap->bufsize / sizeof(struct ostatfs); 643 if (count > SIZE_MAX / sizeof(struct statfs)) 644 return (EINVAL); 645 size = count * sizeof(struct statfs); 646 error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE, 647 uap->mode); 648 if (error == 0) 649 td->td_retval[0] = count; 650 if (size != 0) { 651 sp = buf; 652 while (count != 0 && error == 0) { 653 freebsd4_cvtstatfs(sp, &osb); 654 error = copyout(&osb, uap->buf, sizeof(osb)); 655 sp++; 656 uap->buf++; 657 count--; 658 } 659 free(buf, M_STATFS); 660 } 661 return (error); 662 } 663 664 /* 665 * Implement fstatfs() for (NFS) file handles. 666 */ 667 #ifndef _SYS_SYSPROTO_H_ 668 struct freebsd4_fhstatfs_args { 669 struct fhandle *u_fhp; 670 struct ostatfs *buf; 671 }; 672 #endif 673 int 674 freebsd4_fhstatfs(struct thread *td, struct freebsd4_fhstatfs_args *uap) 675 { 676 struct ostatfs osb; 677 struct statfs *sfp; 678 fhandle_t fh; 679 int error; 680 681 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t)); 682 if (error != 0) 683 return (error); 684 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 685 error = kern_fhstatfs(td, fh, sfp); 686 if (error == 0) { 687 freebsd4_cvtstatfs(sfp, &osb); 688 error = copyout(&osb, uap->buf, sizeof(osb)); 689 } 690 free(sfp, M_STATFS); 691 return (error); 692 } 693 694 /* 695 * Convert a new format statfs structure to an old format statfs structure. 696 */ 697 static void 698 freebsd4_cvtstatfs(struct statfs *nsp, struct ostatfs *osp) 699 { 700 701 statfs_scale_blocks(nsp, LONG_MAX); 702 bzero(osp, sizeof(*osp)); 703 osp->f_bsize = nsp->f_bsize; 704 osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX); 705 osp->f_blocks = nsp->f_blocks; 706 osp->f_bfree = nsp->f_bfree; 707 osp->f_bavail = nsp->f_bavail; 708 osp->f_files = MIN(nsp->f_files, LONG_MAX); 709 osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX); 710 osp->f_owner = nsp->f_owner; 711 osp->f_type = nsp->f_type; 712 osp->f_flags = nsp->f_flags; 713 osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX); 714 osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX); 715 osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX); 716 osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX); 717 strlcpy(osp->f_fstypename, nsp->f_fstypename, 718 MIN(MFSNAMELEN, OMFSNAMELEN)); 719 strlcpy(osp->f_mntonname, nsp->f_mntonname, 720 MIN(MNAMELEN, OMNAMELEN)); 721 strlcpy(osp->f_mntfromname, nsp->f_mntfromname, 722 MIN(MNAMELEN, OMNAMELEN)); 723 osp->f_fsid = nsp->f_fsid; 724 } 725 #endif /* COMPAT_FREEBSD4 */ 726 727 #if defined(COMPAT_FREEBSD11) 728 /* 729 * Get old format filesystem statistics. 730 */ 731 static void freebsd11_cvtstatfs(struct statfs *, struct freebsd11_statfs *); 732 733 int 734 freebsd11_statfs(struct thread *td, struct freebsd11_statfs_args *uap) 735 { 736 struct freebsd11_statfs osb; 737 struct statfs *sfp; 738 int error; 739 740 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 741 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp); 742 if (error == 0) { 743 freebsd11_cvtstatfs(sfp, &osb); 744 error = copyout(&osb, uap->buf, sizeof(osb)); 745 } 746 free(sfp, M_STATFS); 747 return (error); 748 } 749 750 /* 751 * Get filesystem statistics. 752 */ 753 int 754 freebsd11_fstatfs(struct thread *td, struct freebsd11_fstatfs_args *uap) 755 { 756 struct freebsd11_statfs osb; 757 struct statfs *sfp; 758 int error; 759 760 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 761 error = kern_fstatfs(td, uap->fd, sfp); 762 if (error == 0) { 763 freebsd11_cvtstatfs(sfp, &osb); 764 error = copyout(&osb, uap->buf, sizeof(osb)); 765 } 766 free(sfp, M_STATFS); 767 return (error); 768 } 769 770 /* 771 * Get statistics on all filesystems. 772 */ 773 int 774 freebsd11_getfsstat(struct thread *td, struct freebsd11_getfsstat_args *uap) 775 { 776 struct freebsd11_statfs osb; 777 struct statfs *buf, *sp; 778 size_t count, size; 779 int error; 780 781 count = uap->bufsize / sizeof(struct ostatfs); 782 size = count * sizeof(struct statfs); 783 error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE, 784 uap->mode); 785 if (error == 0) 786 td->td_retval[0] = count; 787 if (size > 0) { 788 sp = buf; 789 while (count > 0 && error == 0) { 790 freebsd11_cvtstatfs(sp, &osb); 791 error = copyout(&osb, uap->buf, sizeof(osb)); 792 sp++; 793 uap->buf++; 794 count--; 795 } 796 free(buf, M_STATFS); 797 } 798 return (error); 799 } 800 801 /* 802 * Implement fstatfs() for (NFS) file handles. 803 */ 804 int 805 freebsd11_fhstatfs(struct thread *td, struct freebsd11_fhstatfs_args *uap) 806 { 807 struct freebsd11_statfs osb; 808 struct statfs *sfp; 809 fhandle_t fh; 810 int error; 811 812 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t)); 813 if (error) 814 return (error); 815 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 816 error = kern_fhstatfs(td, fh, sfp); 817 if (error == 0) { 818 freebsd11_cvtstatfs(sfp, &osb); 819 error = copyout(&osb, uap->buf, sizeof(osb)); 820 } 821 free(sfp, M_STATFS); 822 return (error); 823 } 824 825 /* 826 * Convert a new format statfs structure to an old format statfs structure. 827 */ 828 static void 829 freebsd11_cvtstatfs(struct statfs *nsp, struct freebsd11_statfs *osp) 830 { 831 832 bzero(osp, sizeof(*osp)); 833 osp->f_version = FREEBSD11_STATFS_VERSION; 834 osp->f_type = nsp->f_type; 835 osp->f_flags = nsp->f_flags; 836 osp->f_bsize = nsp->f_bsize; 837 osp->f_iosize = nsp->f_iosize; 838 osp->f_blocks = nsp->f_blocks; 839 osp->f_bfree = nsp->f_bfree; 840 osp->f_bavail = nsp->f_bavail; 841 osp->f_files = nsp->f_files; 842 osp->f_ffree = nsp->f_ffree; 843 osp->f_syncwrites = nsp->f_syncwrites; 844 osp->f_asyncwrites = nsp->f_asyncwrites; 845 osp->f_syncreads = nsp->f_syncreads; 846 osp->f_asyncreads = nsp->f_asyncreads; 847 osp->f_namemax = nsp->f_namemax; 848 osp->f_owner = nsp->f_owner; 849 osp->f_fsid = nsp->f_fsid; 850 strlcpy(osp->f_fstypename, nsp->f_fstypename, 851 MIN(MFSNAMELEN, sizeof(osp->f_fstypename))); 852 strlcpy(osp->f_mntonname, nsp->f_mntonname, 853 MIN(MNAMELEN, sizeof(osp->f_mntonname))); 854 strlcpy(osp->f_mntfromname, nsp->f_mntfromname, 855 MIN(MNAMELEN, sizeof(osp->f_mntfromname))); 856 } 857 #endif /* COMPAT_FREEBSD11 */ 858 859 /* 860 * Change current working directory to a given file descriptor. 861 */ 862 #ifndef _SYS_SYSPROTO_H_ 863 struct fchdir_args { 864 int fd; 865 }; 866 #endif 867 int 868 sys_fchdir(struct thread *td, struct fchdir_args *uap) 869 { 870 struct vnode *vp, *tdp; 871 struct mount *mp; 872 struct file *fp; 873 int error; 874 875 AUDIT_ARG_FD(uap->fd); 876 error = getvnode(td, uap->fd, &cap_fchdir_rights, 877 &fp); 878 if (error != 0) 879 return (error); 880 vp = fp->f_vnode; 881 vrefact(vp); 882 fdrop(fp, td); 883 vn_lock(vp, LK_SHARED | LK_RETRY); 884 AUDIT_ARG_VNODE1(vp); 885 error = change_dir(vp, td); 886 while (!error && (mp = vp->v_mountedhere) != NULL) { 887 if (vfs_busy(mp, 0)) 888 continue; 889 error = VFS_ROOT(mp, LK_SHARED, &tdp); 890 vfs_unbusy(mp); 891 if (error != 0) 892 break; 893 vput(vp); 894 vp = tdp; 895 } 896 if (error != 0) { 897 vput(vp); 898 return (error); 899 } 900 VOP_UNLOCK(vp); 901 pwd_chdir(td, vp); 902 return (0); 903 } 904 905 /* 906 * Change current working directory (``.''). 907 */ 908 #ifndef _SYS_SYSPROTO_H_ 909 struct chdir_args { 910 char *path; 911 }; 912 #endif 913 int 914 sys_chdir(struct thread *td, struct chdir_args *uap) 915 { 916 917 return (kern_chdir(td, uap->path, UIO_USERSPACE)); 918 } 919 920 int 921 kern_chdir(struct thread *td, const char *path, enum uio_seg pathseg) 922 { 923 struct nameidata nd; 924 int error; 925 926 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1, 927 pathseg, path, td); 928 if ((error = namei(&nd)) != 0) 929 return (error); 930 if ((error = change_dir(nd.ni_vp, td)) != 0) { 931 vput(nd.ni_vp); 932 NDFREE(&nd, NDF_ONLY_PNBUF); 933 return (error); 934 } 935 VOP_UNLOCK(nd.ni_vp); 936 NDFREE(&nd, NDF_ONLY_PNBUF); 937 pwd_chdir(td, nd.ni_vp); 938 return (0); 939 } 940 941 /* 942 * Change notion of root (``/'') directory. 943 */ 944 #ifndef _SYS_SYSPROTO_H_ 945 struct chroot_args { 946 char *path; 947 }; 948 #endif 949 int 950 sys_chroot(struct thread *td, struct chroot_args *uap) 951 { 952 struct nameidata nd; 953 int error; 954 955 error = priv_check(td, PRIV_VFS_CHROOT); 956 if (error != 0) 957 return (error); 958 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1, 959 UIO_USERSPACE, uap->path, td); 960 error = namei(&nd); 961 if (error != 0) 962 goto error; 963 error = change_dir(nd.ni_vp, td); 964 if (error != 0) 965 goto e_vunlock; 966 #ifdef MAC 967 error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp); 968 if (error != 0) 969 goto e_vunlock; 970 #endif 971 VOP_UNLOCK(nd.ni_vp); 972 error = pwd_chroot(td, nd.ni_vp); 973 vrele(nd.ni_vp); 974 NDFREE(&nd, NDF_ONLY_PNBUF); 975 return (error); 976 e_vunlock: 977 vput(nd.ni_vp); 978 error: 979 NDFREE(&nd, NDF_ONLY_PNBUF); 980 return (error); 981 } 982 983 /* 984 * Common routine for chroot and chdir. Callers must provide a locked vnode 985 * instance. 986 */ 987 int 988 change_dir(struct vnode *vp, struct thread *td) 989 { 990 #ifdef MAC 991 int error; 992 #endif 993 994 ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked"); 995 if (vp->v_type != VDIR) 996 return (ENOTDIR); 997 #ifdef MAC 998 error = mac_vnode_check_chdir(td->td_ucred, vp); 999 if (error != 0) 1000 return (error); 1001 #endif 1002 return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td)); 1003 } 1004 1005 static __inline void 1006 flags_to_rights(int flags, cap_rights_t *rightsp) 1007 { 1008 1009 if (flags & O_EXEC) { 1010 cap_rights_set_one(rightsp, CAP_FEXECVE); 1011 } else { 1012 switch ((flags & O_ACCMODE)) { 1013 case O_RDONLY: 1014 cap_rights_set_one(rightsp, CAP_READ); 1015 break; 1016 case O_RDWR: 1017 cap_rights_set_one(rightsp, CAP_READ); 1018 /* FALLTHROUGH */ 1019 case O_WRONLY: 1020 cap_rights_set_one(rightsp, CAP_WRITE); 1021 if (!(flags & (O_APPEND | O_TRUNC))) 1022 cap_rights_set_one(rightsp, CAP_SEEK); 1023 break; 1024 } 1025 } 1026 1027 if (flags & O_CREAT) 1028 cap_rights_set_one(rightsp, CAP_CREATE); 1029 1030 if (flags & O_TRUNC) 1031 cap_rights_set_one(rightsp, CAP_FTRUNCATE); 1032 1033 if (flags & (O_SYNC | O_FSYNC)) 1034 cap_rights_set_one(rightsp, CAP_FSYNC); 1035 1036 if (flags & (O_EXLOCK | O_SHLOCK)) 1037 cap_rights_set_one(rightsp, CAP_FLOCK); 1038 } 1039 1040 /* 1041 * Check permissions, allocate an open file structure, and call the device 1042 * open routine if any. 1043 */ 1044 #ifndef _SYS_SYSPROTO_H_ 1045 struct open_args { 1046 char *path; 1047 int flags; 1048 int mode; 1049 }; 1050 #endif 1051 int 1052 sys_open(struct thread *td, struct open_args *uap) 1053 { 1054 1055 return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 1056 uap->flags, uap->mode)); 1057 } 1058 1059 #ifndef _SYS_SYSPROTO_H_ 1060 struct openat_args { 1061 int fd; 1062 char *path; 1063 int flag; 1064 int mode; 1065 }; 1066 #endif 1067 int 1068 sys_openat(struct thread *td, struct openat_args *uap) 1069 { 1070 1071 AUDIT_ARG_FD(uap->fd); 1072 return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag, 1073 uap->mode)); 1074 } 1075 1076 int 1077 kern_openat(struct thread *td, int fd, const char *path, enum uio_seg pathseg, 1078 int flags, int mode) 1079 { 1080 struct proc *p = td->td_proc; 1081 struct filedesc *fdp = p->p_fd; 1082 struct file *fp; 1083 struct vnode *vp; 1084 struct nameidata nd; 1085 cap_rights_t rights; 1086 int cmode, error, indx; 1087 1088 indx = -1; 1089 1090 AUDIT_ARG_FFLAGS(flags); 1091 AUDIT_ARG_MODE(mode); 1092 cap_rights_init_one(&rights, CAP_LOOKUP); 1093 flags_to_rights(flags, &rights); 1094 /* 1095 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags 1096 * may be specified. 1097 */ 1098 if (flags & O_EXEC) { 1099 if (flags & O_ACCMODE) 1100 return (EINVAL); 1101 } else if ((flags & O_ACCMODE) == O_ACCMODE) { 1102 return (EINVAL); 1103 } else { 1104 flags = FFLAGS(flags); 1105 } 1106 1107 /* 1108 * Allocate a file structure. The descriptor to reference it 1109 * is allocated and set by finstall() below. 1110 */ 1111 error = falloc_noinstall(td, &fp); 1112 if (error != 0) 1113 return (error); 1114 /* 1115 * An extra reference on `fp' has been held for us by 1116 * falloc_noinstall(). 1117 */ 1118 /* Set the flags early so the finit in devfs can pick them up. */ 1119 fp->f_flag = flags & FMASK; 1120 cmode = ((mode & ~fdp->fd_cmask) & ALLPERMS) & ~S_ISTXT; 1121 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd, 1122 &rights, td); 1123 td->td_dupfd = -1; /* XXX check for fdopen */ 1124 error = vn_open(&nd, &flags, cmode, fp); 1125 if (error != 0) { 1126 /* 1127 * If the vn_open replaced the method vector, something 1128 * wonderous happened deep below and we just pass it up 1129 * pretending we know what we do. 1130 */ 1131 if (error == ENXIO && fp->f_ops != &badfileops) 1132 goto success; 1133 1134 /* 1135 * Handle special fdopen() case. bleh. 1136 * 1137 * Don't do this for relative (capability) lookups; we don't 1138 * understand exactly what would happen, and we don't think 1139 * that it ever should. 1140 */ 1141 if ((nd.ni_lcf & NI_LCF_STRICTRELATIVE) == 0 && 1142 (error == ENODEV || error == ENXIO) && 1143 td->td_dupfd >= 0) { 1144 error = dupfdopen(td, fdp, td->td_dupfd, flags, error, 1145 &indx); 1146 if (error == 0) 1147 goto success; 1148 } 1149 1150 goto bad; 1151 } 1152 td->td_dupfd = 0; 1153 NDFREE(&nd, NDF_ONLY_PNBUF); 1154 vp = nd.ni_vp; 1155 1156 /* 1157 * Store the vnode, for any f_type. Typically, the vnode use 1158 * count is decremented by direct call to vn_closefile() for 1159 * files that switched type in the cdevsw fdopen() method. 1160 */ 1161 fp->f_vnode = vp; 1162 /* 1163 * If the file wasn't claimed by devfs bind it to the normal 1164 * vnode operations here. 1165 */ 1166 if (fp->f_ops == &badfileops) { 1167 KASSERT(vp->v_type != VFIFO, ("Unexpected fifo.")); 1168 fp->f_seqcount[UIO_READ] = 1; 1169 fp->f_seqcount[UIO_WRITE] = 1; 1170 finit(fp, (flags & FMASK) | (fp->f_flag & FHASLOCK), 1171 DTYPE_VNODE, vp, &vnops); 1172 } 1173 1174 VOP_UNLOCK(vp); 1175 if (flags & O_TRUNC) { 1176 error = fo_truncate(fp, 0, td->td_ucred, td); 1177 if (error != 0) 1178 goto bad; 1179 } 1180 success: 1181 /* 1182 * If we haven't already installed the FD (for dupfdopen), do so now. 1183 */ 1184 if (indx == -1) { 1185 struct filecaps *fcaps; 1186 1187 #ifdef CAPABILITIES 1188 if ((nd.ni_lcf & NI_LCF_STRICTRELATIVE) != 0) 1189 fcaps = &nd.ni_filecaps; 1190 else 1191 #endif 1192 fcaps = NULL; 1193 error = finstall(td, fp, &indx, flags, fcaps); 1194 /* On success finstall() consumes fcaps. */ 1195 if (error != 0) { 1196 filecaps_free(&nd.ni_filecaps); 1197 goto bad; 1198 } 1199 } else { 1200 filecaps_free(&nd.ni_filecaps); 1201 } 1202 1203 /* 1204 * Release our private reference, leaving the one associated with 1205 * the descriptor table intact. 1206 */ 1207 fdrop(fp, td); 1208 td->td_retval[0] = indx; 1209 return (0); 1210 bad: 1211 KASSERT(indx == -1, ("indx=%d, should be -1", indx)); 1212 fdrop(fp, td); 1213 return (error); 1214 } 1215 1216 #ifdef COMPAT_43 1217 /* 1218 * Create a file. 1219 */ 1220 #ifndef _SYS_SYSPROTO_H_ 1221 struct ocreat_args { 1222 char *path; 1223 int mode; 1224 }; 1225 #endif 1226 int 1227 ocreat(struct thread *td, struct ocreat_args *uap) 1228 { 1229 1230 return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 1231 O_WRONLY | O_CREAT | O_TRUNC, uap->mode)); 1232 } 1233 #endif /* COMPAT_43 */ 1234 1235 /* 1236 * Create a special file. 1237 */ 1238 #ifndef _SYS_SYSPROTO_H_ 1239 struct mknodat_args { 1240 int fd; 1241 char *path; 1242 mode_t mode; 1243 dev_t dev; 1244 }; 1245 #endif 1246 int 1247 sys_mknodat(struct thread *td, struct mknodat_args *uap) 1248 { 1249 1250 return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode, 1251 uap->dev)); 1252 } 1253 1254 #if defined(COMPAT_FREEBSD11) 1255 int 1256 freebsd11_mknod(struct thread *td, 1257 struct freebsd11_mknod_args *uap) 1258 { 1259 1260 return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 1261 uap->mode, uap->dev)); 1262 } 1263 1264 int 1265 freebsd11_mknodat(struct thread *td, 1266 struct freebsd11_mknodat_args *uap) 1267 { 1268 1269 return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode, 1270 uap->dev)); 1271 } 1272 #endif /* COMPAT_FREEBSD11 */ 1273 1274 int 1275 kern_mknodat(struct thread *td, int fd, const char *path, enum uio_seg pathseg, 1276 int mode, dev_t dev) 1277 { 1278 struct vnode *vp; 1279 struct mount *mp; 1280 struct vattr vattr; 1281 struct nameidata nd; 1282 int error, whiteout = 0; 1283 1284 AUDIT_ARG_MODE(mode); 1285 AUDIT_ARG_DEV(dev); 1286 switch (mode & S_IFMT) { 1287 case S_IFCHR: 1288 case S_IFBLK: 1289 error = priv_check(td, PRIV_VFS_MKNOD_DEV); 1290 if (error == 0 && dev == VNOVAL) 1291 error = EINVAL; 1292 break; 1293 case S_IFWHT: 1294 error = priv_check(td, PRIV_VFS_MKNOD_WHT); 1295 break; 1296 case S_IFIFO: 1297 if (dev == 0) 1298 return (kern_mkfifoat(td, fd, path, pathseg, mode)); 1299 /* FALLTHROUGH */ 1300 default: 1301 error = EINVAL; 1302 break; 1303 } 1304 if (error != 0) 1305 return (error); 1306 restart: 1307 bwillwrite(); 1308 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 | 1309 NOCACHE, pathseg, path, fd, &cap_mknodat_rights, 1310 td); 1311 if ((error = namei(&nd)) != 0) 1312 return (error); 1313 vp = nd.ni_vp; 1314 if (vp != NULL) { 1315 NDFREE(&nd, NDF_ONLY_PNBUF); 1316 if (vp == nd.ni_dvp) 1317 vrele(nd.ni_dvp); 1318 else 1319 vput(nd.ni_dvp); 1320 vrele(vp); 1321 return (EEXIST); 1322 } else { 1323 VATTR_NULL(&vattr); 1324 vattr.va_mode = (mode & ALLPERMS) & 1325 ~td->td_proc->p_fd->fd_cmask; 1326 vattr.va_rdev = dev; 1327 whiteout = 0; 1328 1329 switch (mode & S_IFMT) { 1330 case S_IFCHR: 1331 vattr.va_type = VCHR; 1332 break; 1333 case S_IFBLK: 1334 vattr.va_type = VBLK; 1335 break; 1336 case S_IFWHT: 1337 whiteout = 1; 1338 break; 1339 default: 1340 panic("kern_mknod: invalid mode"); 1341 } 1342 } 1343 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1344 NDFREE(&nd, NDF_ONLY_PNBUF); 1345 vput(nd.ni_dvp); 1346 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0) 1347 return (error); 1348 goto restart; 1349 } 1350 #ifdef MAC 1351 if (error == 0 && !whiteout) 1352 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, 1353 &nd.ni_cnd, &vattr); 1354 #endif 1355 if (error == 0) { 1356 if (whiteout) 1357 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1358 else { 1359 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1360 &nd.ni_cnd, &vattr); 1361 if (error == 0) 1362 vput(nd.ni_vp); 1363 } 1364 } 1365 NDFREE(&nd, NDF_ONLY_PNBUF); 1366 vput(nd.ni_dvp); 1367 vn_finished_write(mp); 1368 return (error); 1369 } 1370 1371 /* 1372 * Create a named pipe. 1373 */ 1374 #ifndef _SYS_SYSPROTO_H_ 1375 struct mkfifo_args { 1376 char *path; 1377 int mode; 1378 }; 1379 #endif 1380 int 1381 sys_mkfifo(struct thread *td, struct mkfifo_args *uap) 1382 { 1383 1384 return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 1385 uap->mode)); 1386 } 1387 1388 #ifndef _SYS_SYSPROTO_H_ 1389 struct mkfifoat_args { 1390 int fd; 1391 char *path; 1392 mode_t mode; 1393 }; 1394 #endif 1395 int 1396 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap) 1397 { 1398 1399 return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE, 1400 uap->mode)); 1401 } 1402 1403 int 1404 kern_mkfifoat(struct thread *td, int fd, const char *path, 1405 enum uio_seg pathseg, int mode) 1406 { 1407 struct mount *mp; 1408 struct vattr vattr; 1409 struct nameidata nd; 1410 int error; 1411 1412 AUDIT_ARG_MODE(mode); 1413 restart: 1414 bwillwrite(); 1415 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 | 1416 NOCACHE, pathseg, path, fd, &cap_mkfifoat_rights, 1417 td); 1418 if ((error = namei(&nd)) != 0) 1419 return (error); 1420 if (nd.ni_vp != NULL) { 1421 NDFREE(&nd, NDF_ONLY_PNBUF); 1422 if (nd.ni_vp == nd.ni_dvp) 1423 vrele(nd.ni_dvp); 1424 else 1425 vput(nd.ni_dvp); 1426 vrele(nd.ni_vp); 1427 return (EEXIST); 1428 } 1429 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1430 NDFREE(&nd, NDF_ONLY_PNBUF); 1431 vput(nd.ni_dvp); 1432 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0) 1433 return (error); 1434 goto restart; 1435 } 1436 VATTR_NULL(&vattr); 1437 vattr.va_type = VFIFO; 1438 vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask; 1439 #ifdef MAC 1440 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd, 1441 &vattr); 1442 if (error != 0) 1443 goto out; 1444 #endif 1445 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1446 if (error == 0) 1447 vput(nd.ni_vp); 1448 #ifdef MAC 1449 out: 1450 #endif 1451 vput(nd.ni_dvp); 1452 vn_finished_write(mp); 1453 NDFREE(&nd, NDF_ONLY_PNBUF); 1454 return (error); 1455 } 1456 1457 /* 1458 * Make a hard file link. 1459 */ 1460 #ifndef _SYS_SYSPROTO_H_ 1461 struct link_args { 1462 char *path; 1463 char *link; 1464 }; 1465 #endif 1466 int 1467 sys_link(struct thread *td, struct link_args *uap) 1468 { 1469 1470 return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link, 1471 UIO_USERSPACE, FOLLOW)); 1472 } 1473 1474 #ifndef _SYS_SYSPROTO_H_ 1475 struct linkat_args { 1476 int fd1; 1477 char *path1; 1478 int fd2; 1479 char *path2; 1480 int flag; 1481 }; 1482 #endif 1483 int 1484 sys_linkat(struct thread *td, struct linkat_args *uap) 1485 { 1486 int flag; 1487 1488 flag = uap->flag; 1489 if ((flag & ~(AT_SYMLINK_FOLLOW | AT_BENEATH)) != 0) 1490 return (EINVAL); 1491 1492 return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2, 1493 UIO_USERSPACE, ((flag & AT_SYMLINK_FOLLOW) != 0 ? FOLLOW : 1494 NOFOLLOW) | ((flag & AT_BENEATH) != 0 ? BENEATH : 0))); 1495 } 1496 1497 int hardlink_check_uid = 0; 1498 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW, 1499 &hardlink_check_uid, 0, 1500 "Unprivileged processes cannot create hard links to files owned by other " 1501 "users"); 1502 static int hardlink_check_gid = 0; 1503 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW, 1504 &hardlink_check_gid, 0, 1505 "Unprivileged processes cannot create hard links to files owned by other " 1506 "groups"); 1507 1508 static int 1509 can_hardlink(struct vnode *vp, struct ucred *cred) 1510 { 1511 struct vattr va; 1512 int error; 1513 1514 if (!hardlink_check_uid && !hardlink_check_gid) 1515 return (0); 1516 1517 error = VOP_GETATTR(vp, &va, cred); 1518 if (error != 0) 1519 return (error); 1520 1521 if (hardlink_check_uid && cred->cr_uid != va.va_uid) { 1522 error = priv_check_cred(cred, PRIV_VFS_LINK); 1523 if (error != 0) 1524 return (error); 1525 } 1526 1527 if (hardlink_check_gid && !groupmember(va.va_gid, cred)) { 1528 error = priv_check_cred(cred, PRIV_VFS_LINK); 1529 if (error != 0) 1530 return (error); 1531 } 1532 1533 return (0); 1534 } 1535 1536 int 1537 kern_linkat(struct thread *td, int fd1, int fd2, const char *path1, 1538 const char *path2, enum uio_seg segflag, int follow) 1539 { 1540 struct nameidata nd; 1541 int error; 1542 1543 do { 1544 bwillwrite(); 1545 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, segflag, 1546 path1, fd1, &cap_linkat_source_rights, td); 1547 if ((error = namei(&nd)) != 0) 1548 return (error); 1549 NDFREE(&nd, NDF_ONLY_PNBUF); 1550 error = kern_linkat_vp(td, nd.ni_vp, fd2, path2, segflag); 1551 } while (error == EAGAIN); 1552 return (error); 1553 } 1554 1555 static int 1556 kern_linkat_vp(struct thread *td, struct vnode *vp, int fd, const char *path, 1557 enum uio_seg segflag) 1558 { 1559 struct nameidata nd; 1560 struct mount *mp; 1561 int error; 1562 1563 if (vp->v_type == VDIR) { 1564 vrele(vp); 1565 return (EPERM); /* POSIX */ 1566 } 1567 NDINIT_ATRIGHTS(&nd, CREATE, 1568 LOCKPARENT | SAVENAME | AUDITVNODE2 | NOCACHE, segflag, path, fd, 1569 &cap_linkat_target_rights, td); 1570 if ((error = namei(&nd)) == 0) { 1571 if (nd.ni_vp != NULL) { 1572 NDFREE(&nd, NDF_ONLY_PNBUF); 1573 if (nd.ni_dvp == nd.ni_vp) 1574 vrele(nd.ni_dvp); 1575 else 1576 vput(nd.ni_dvp); 1577 vrele(nd.ni_vp); 1578 vrele(vp); 1579 return (EEXIST); 1580 } else if (nd.ni_dvp->v_mount != vp->v_mount) { 1581 /* 1582 * Cross-device link. No need to recheck 1583 * vp->v_type, since it cannot change, except 1584 * to VBAD. 1585 */ 1586 NDFREE(&nd, NDF_ONLY_PNBUF); 1587 vput(nd.ni_dvp); 1588 vrele(vp); 1589 return (EXDEV); 1590 } else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) { 1591 error = can_hardlink(vp, td->td_ucred); 1592 #ifdef MAC 1593 if (error == 0) 1594 error = mac_vnode_check_link(td->td_ucred, 1595 nd.ni_dvp, vp, &nd.ni_cnd); 1596 #endif 1597 if (error != 0) { 1598 vput(vp); 1599 vput(nd.ni_dvp); 1600 NDFREE(&nd, NDF_ONLY_PNBUF); 1601 return (error); 1602 } 1603 error = vn_start_write(vp, &mp, V_NOWAIT); 1604 if (error != 0) { 1605 vput(vp); 1606 vput(nd.ni_dvp); 1607 NDFREE(&nd, NDF_ONLY_PNBUF); 1608 error = vn_start_write(NULL, &mp, 1609 V_XSLEEP | PCATCH); 1610 if (error != 0) 1611 return (error); 1612 return (EAGAIN); 1613 } 1614 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1615 VOP_UNLOCK(vp); 1616 vput(nd.ni_dvp); 1617 vn_finished_write(mp); 1618 NDFREE(&nd, NDF_ONLY_PNBUF); 1619 } else { 1620 vput(nd.ni_dvp); 1621 NDFREE(&nd, NDF_ONLY_PNBUF); 1622 vrele(vp); 1623 return (EAGAIN); 1624 } 1625 } 1626 vrele(vp); 1627 return (error); 1628 } 1629 1630 /* 1631 * Make a symbolic link. 1632 */ 1633 #ifndef _SYS_SYSPROTO_H_ 1634 struct symlink_args { 1635 char *path; 1636 char *link; 1637 }; 1638 #endif 1639 int 1640 sys_symlink(struct thread *td, struct symlink_args *uap) 1641 { 1642 1643 return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link, 1644 UIO_USERSPACE)); 1645 } 1646 1647 #ifndef _SYS_SYSPROTO_H_ 1648 struct symlinkat_args { 1649 char *path; 1650 int fd; 1651 char *path2; 1652 }; 1653 #endif 1654 int 1655 sys_symlinkat(struct thread *td, struct symlinkat_args *uap) 1656 { 1657 1658 return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2, 1659 UIO_USERSPACE)); 1660 } 1661 1662 int 1663 kern_symlinkat(struct thread *td, const char *path1, int fd, const char *path2, 1664 enum uio_seg segflg) 1665 { 1666 struct mount *mp; 1667 struct vattr vattr; 1668 const char *syspath; 1669 char *tmppath; 1670 struct nameidata nd; 1671 int error; 1672 1673 if (segflg == UIO_SYSSPACE) { 1674 syspath = path1; 1675 } else { 1676 tmppath = uma_zalloc(namei_zone, M_WAITOK); 1677 if ((error = copyinstr(path1, tmppath, MAXPATHLEN, NULL)) != 0) 1678 goto out; 1679 syspath = tmppath; 1680 } 1681 AUDIT_ARG_TEXT(syspath); 1682 restart: 1683 bwillwrite(); 1684 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 | 1685 NOCACHE, segflg, path2, fd, &cap_symlinkat_rights, 1686 td); 1687 if ((error = namei(&nd)) != 0) 1688 goto out; 1689 if (nd.ni_vp) { 1690 NDFREE(&nd, NDF_ONLY_PNBUF); 1691 if (nd.ni_vp == nd.ni_dvp) 1692 vrele(nd.ni_dvp); 1693 else 1694 vput(nd.ni_dvp); 1695 vrele(nd.ni_vp); 1696 error = EEXIST; 1697 goto out; 1698 } 1699 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1700 NDFREE(&nd, NDF_ONLY_PNBUF); 1701 vput(nd.ni_dvp); 1702 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0) 1703 goto out; 1704 goto restart; 1705 } 1706 VATTR_NULL(&vattr); 1707 vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask; 1708 #ifdef MAC 1709 vattr.va_type = VLNK; 1710 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd, 1711 &vattr); 1712 if (error != 0) 1713 goto out2; 1714 #endif 1715 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath); 1716 if (error == 0) 1717 vput(nd.ni_vp); 1718 #ifdef MAC 1719 out2: 1720 #endif 1721 NDFREE(&nd, NDF_ONLY_PNBUF); 1722 vput(nd.ni_dvp); 1723 vn_finished_write(mp); 1724 out: 1725 if (segflg != UIO_SYSSPACE) 1726 uma_zfree(namei_zone, tmppath); 1727 return (error); 1728 } 1729 1730 /* 1731 * Delete a whiteout from the filesystem. 1732 */ 1733 #ifndef _SYS_SYSPROTO_H_ 1734 struct undelete_args { 1735 char *path; 1736 }; 1737 #endif 1738 int 1739 sys_undelete(struct thread *td, struct undelete_args *uap) 1740 { 1741 struct mount *mp; 1742 struct nameidata nd; 1743 int error; 1744 1745 restart: 1746 bwillwrite(); 1747 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1, 1748 UIO_USERSPACE, uap->path, td); 1749 error = namei(&nd); 1750 if (error != 0) 1751 return (error); 1752 1753 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 1754 NDFREE(&nd, NDF_ONLY_PNBUF); 1755 if (nd.ni_vp == nd.ni_dvp) 1756 vrele(nd.ni_dvp); 1757 else 1758 vput(nd.ni_dvp); 1759 if (nd.ni_vp) 1760 vrele(nd.ni_vp); 1761 return (EEXIST); 1762 } 1763 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1764 NDFREE(&nd, NDF_ONLY_PNBUF); 1765 vput(nd.ni_dvp); 1766 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0) 1767 return (error); 1768 goto restart; 1769 } 1770 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE); 1771 NDFREE(&nd, NDF_ONLY_PNBUF); 1772 vput(nd.ni_dvp); 1773 vn_finished_write(mp); 1774 return (error); 1775 } 1776 1777 /* 1778 * Delete a name from the filesystem. 1779 */ 1780 #ifndef _SYS_SYSPROTO_H_ 1781 struct unlink_args { 1782 char *path; 1783 }; 1784 #endif 1785 int 1786 sys_unlink(struct thread *td, struct unlink_args *uap) 1787 { 1788 1789 return (kern_funlinkat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE, 1790 0, 0)); 1791 } 1792 1793 static int 1794 kern_funlinkat_ex(struct thread *td, int dfd, const char *path, int fd, 1795 int flag, enum uio_seg pathseg, ino_t oldinum) 1796 { 1797 1798 if ((flag & ~AT_REMOVEDIR) != 0) 1799 return (EINVAL); 1800 1801 if ((flag & AT_REMOVEDIR) != 0) 1802 return (kern_frmdirat(td, dfd, path, fd, UIO_USERSPACE, 0)); 1803 1804 return (kern_funlinkat(td, dfd, path, fd, UIO_USERSPACE, 0, 0)); 1805 } 1806 1807 #ifndef _SYS_SYSPROTO_H_ 1808 struct unlinkat_args { 1809 int fd; 1810 char *path; 1811 int flag; 1812 }; 1813 #endif 1814 int 1815 sys_unlinkat(struct thread *td, struct unlinkat_args *uap) 1816 { 1817 1818 return (kern_funlinkat_ex(td, uap->fd, uap->path, FD_NONE, uap->flag, 1819 UIO_USERSPACE, 0)); 1820 } 1821 1822 #ifndef _SYS_SYSPROTO_H_ 1823 struct funlinkat_args { 1824 int dfd; 1825 const char *path; 1826 int fd; 1827 int flag; 1828 }; 1829 #endif 1830 int 1831 sys_funlinkat(struct thread *td, struct funlinkat_args *uap) 1832 { 1833 1834 return (kern_funlinkat_ex(td, uap->dfd, uap->path, uap->fd, uap->flag, 1835 UIO_USERSPACE, 0)); 1836 } 1837 1838 int 1839 kern_funlinkat(struct thread *td, int dfd, const char *path, int fd, 1840 enum uio_seg pathseg, int flag, ino_t oldinum) 1841 { 1842 struct mount *mp; 1843 struct file *fp; 1844 struct vnode *vp; 1845 struct nameidata nd; 1846 struct stat sb; 1847 int error; 1848 1849 fp = NULL; 1850 if (fd != FD_NONE) { 1851 error = getvnode(td, fd, &cap_no_rights, &fp); 1852 if (error != 0) 1853 return (error); 1854 } 1855 1856 restart: 1857 bwillwrite(); 1858 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 | 1859 ((flag & AT_BENEATH) != 0 ? BENEATH : 0), 1860 pathseg, path, dfd, &cap_unlinkat_rights, td); 1861 if ((error = namei(&nd)) != 0) { 1862 if (error == EINVAL) 1863 error = EPERM; 1864 goto fdout; 1865 } 1866 vp = nd.ni_vp; 1867 if (vp->v_type == VDIR && oldinum == 0) { 1868 error = EPERM; /* POSIX */ 1869 } else if (oldinum != 0 && 1870 ((error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td)) == 0) && 1871 sb.st_ino != oldinum) { 1872 error = EIDRM; /* Identifier removed */ 1873 } else if (fp != NULL && fp->f_vnode != vp) { 1874 if (VN_IS_DOOMED(fp->f_vnode)) 1875 error = EBADF; 1876 else 1877 error = EDEADLK; 1878 } else { 1879 /* 1880 * The root of a mounted filesystem cannot be deleted. 1881 * 1882 * XXX: can this only be a VDIR case? 1883 */ 1884 if (vp->v_vflag & VV_ROOT) 1885 error = EBUSY; 1886 } 1887 if (error == 0) { 1888 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1889 NDFREE(&nd, NDF_ONLY_PNBUF); 1890 vput(nd.ni_dvp); 1891 if (vp == nd.ni_dvp) 1892 vrele(vp); 1893 else 1894 vput(vp); 1895 if ((error = vn_start_write(NULL, &mp, 1896 V_XSLEEP | PCATCH)) != 0) { 1897 goto fdout; 1898 } 1899 goto restart; 1900 } 1901 #ifdef MAC 1902 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp, 1903 &nd.ni_cnd); 1904 if (error != 0) 1905 goto out; 1906 #endif 1907 vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK); 1908 error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd); 1909 #ifdef MAC 1910 out: 1911 #endif 1912 vn_finished_write(mp); 1913 } 1914 NDFREE(&nd, NDF_ONLY_PNBUF); 1915 vput(nd.ni_dvp); 1916 if (vp == nd.ni_dvp) 1917 vrele(vp); 1918 else 1919 vput(vp); 1920 fdout: 1921 if (fp != NULL) 1922 fdrop(fp, td); 1923 return (error); 1924 } 1925 1926 /* 1927 * Reposition read/write file offset. 1928 */ 1929 #ifndef _SYS_SYSPROTO_H_ 1930 struct lseek_args { 1931 int fd; 1932 int pad; 1933 off_t offset; 1934 int whence; 1935 }; 1936 #endif 1937 int 1938 sys_lseek(struct thread *td, struct lseek_args *uap) 1939 { 1940 1941 return (kern_lseek(td, uap->fd, uap->offset, uap->whence)); 1942 } 1943 1944 int 1945 kern_lseek(struct thread *td, int fd, off_t offset, int whence) 1946 { 1947 struct file *fp; 1948 int error; 1949 1950 AUDIT_ARG_FD(fd); 1951 error = fget(td, fd, &cap_seek_rights, &fp); 1952 if (error != 0) 1953 return (error); 1954 error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ? 1955 fo_seek(fp, offset, whence, td) : ESPIPE; 1956 fdrop(fp, td); 1957 return (error); 1958 } 1959 1960 #if defined(COMPAT_43) 1961 /* 1962 * Reposition read/write file offset. 1963 */ 1964 #ifndef _SYS_SYSPROTO_H_ 1965 struct olseek_args { 1966 int fd; 1967 long offset; 1968 int whence; 1969 }; 1970 #endif 1971 int 1972 olseek(struct thread *td, struct olseek_args *uap) 1973 { 1974 1975 return (kern_lseek(td, uap->fd, uap->offset, uap->whence)); 1976 } 1977 #endif /* COMPAT_43 */ 1978 1979 #if defined(COMPAT_FREEBSD6) 1980 /* Version with the 'pad' argument */ 1981 int 1982 freebsd6_lseek(struct thread *td, struct freebsd6_lseek_args *uap) 1983 { 1984 1985 return (kern_lseek(td, uap->fd, uap->offset, uap->whence)); 1986 } 1987 #endif 1988 1989 /* 1990 * Check access permissions using passed credentials. 1991 */ 1992 static int 1993 vn_access(struct vnode *vp, int user_flags, struct ucred *cred, 1994 struct thread *td) 1995 { 1996 accmode_t accmode; 1997 int error; 1998 1999 /* Flags == 0 means only check for existence. */ 2000 if (user_flags == 0) 2001 return (0); 2002 2003 accmode = 0; 2004 if (user_flags & R_OK) 2005 accmode |= VREAD; 2006 if (user_flags & W_OK) 2007 accmode |= VWRITE; 2008 if (user_flags & X_OK) 2009 accmode |= VEXEC; 2010 #ifdef MAC 2011 error = mac_vnode_check_access(cred, vp, accmode); 2012 if (error != 0) 2013 return (error); 2014 #endif 2015 if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0) 2016 error = VOP_ACCESS(vp, accmode, cred, td); 2017 return (error); 2018 } 2019 2020 /* 2021 * Check access permissions using "real" credentials. 2022 */ 2023 #ifndef _SYS_SYSPROTO_H_ 2024 struct access_args { 2025 char *path; 2026 int amode; 2027 }; 2028 #endif 2029 int 2030 sys_access(struct thread *td, struct access_args *uap) 2031 { 2032 2033 return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2034 0, uap->amode)); 2035 } 2036 2037 #ifndef _SYS_SYSPROTO_H_ 2038 struct faccessat_args { 2039 int dirfd; 2040 char *path; 2041 int amode; 2042 int flag; 2043 } 2044 #endif 2045 int 2046 sys_faccessat(struct thread *td, struct faccessat_args *uap) 2047 { 2048 2049 return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag, 2050 uap->amode)); 2051 } 2052 2053 int 2054 kern_accessat(struct thread *td, int fd, const char *path, 2055 enum uio_seg pathseg, int flag, int amode) 2056 { 2057 struct ucred *cred, *usecred; 2058 struct vnode *vp; 2059 struct nameidata nd; 2060 int error; 2061 2062 if ((flag & ~(AT_EACCESS | AT_BENEATH)) != 0) 2063 return (EINVAL); 2064 if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0) 2065 return (EINVAL); 2066 2067 /* 2068 * Create and modify a temporary credential instead of one that 2069 * is potentially shared (if we need one). 2070 */ 2071 cred = td->td_ucred; 2072 if ((flag & AT_EACCESS) == 0 && 2073 ((cred->cr_uid != cred->cr_ruid || 2074 cred->cr_rgid != cred->cr_groups[0]))) { 2075 usecred = crdup(cred); 2076 usecred->cr_uid = cred->cr_ruid; 2077 usecred->cr_groups[0] = cred->cr_rgid; 2078 td->td_ucred = usecred; 2079 } else 2080 usecred = cred; 2081 AUDIT_ARG_VALUE(amode); 2082 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | 2083 AUDITVNODE1 | ((flag & AT_BENEATH) != 0 ? BENEATH : 0), 2084 pathseg, path, fd, &cap_fstat_rights, td); 2085 if ((error = namei(&nd)) != 0) 2086 goto out; 2087 vp = nd.ni_vp; 2088 2089 error = vn_access(vp, amode, usecred, td); 2090 NDFREE(&nd, NDF_ONLY_PNBUF); 2091 vput(vp); 2092 out: 2093 if (usecred != cred) { 2094 td->td_ucred = cred; 2095 crfree(usecred); 2096 } 2097 return (error); 2098 } 2099 2100 /* 2101 * Check access permissions using "effective" credentials. 2102 */ 2103 #ifndef _SYS_SYSPROTO_H_ 2104 struct eaccess_args { 2105 char *path; 2106 int amode; 2107 }; 2108 #endif 2109 int 2110 sys_eaccess(struct thread *td, struct eaccess_args *uap) 2111 { 2112 2113 return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2114 AT_EACCESS, uap->amode)); 2115 } 2116 2117 #if defined(COMPAT_43) 2118 /* 2119 * Get file status; this version follows links. 2120 */ 2121 #ifndef _SYS_SYSPROTO_H_ 2122 struct ostat_args { 2123 char *path; 2124 struct ostat *ub; 2125 }; 2126 #endif 2127 int 2128 ostat(struct thread *td, struct ostat_args *uap) 2129 { 2130 struct stat sb; 2131 struct ostat osb; 2132 int error; 2133 2134 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE, 2135 &sb, NULL); 2136 if (error != 0) 2137 return (error); 2138 cvtstat(&sb, &osb); 2139 return (copyout(&osb, uap->ub, sizeof (osb))); 2140 } 2141 2142 /* 2143 * Get file status; this version does not follow links. 2144 */ 2145 #ifndef _SYS_SYSPROTO_H_ 2146 struct olstat_args { 2147 char *path; 2148 struct ostat *ub; 2149 }; 2150 #endif 2151 int 2152 olstat(struct thread *td, struct olstat_args *uap) 2153 { 2154 struct stat sb; 2155 struct ostat osb; 2156 int error; 2157 2158 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path, 2159 UIO_USERSPACE, &sb, NULL); 2160 if (error != 0) 2161 return (error); 2162 cvtstat(&sb, &osb); 2163 return (copyout(&osb, uap->ub, sizeof (osb))); 2164 } 2165 2166 /* 2167 * Convert from an old to a new stat structure. 2168 * XXX: many values are blindly truncated. 2169 */ 2170 void 2171 cvtstat(struct stat *st, struct ostat *ost) 2172 { 2173 2174 bzero(ost, sizeof(*ost)); 2175 ost->st_dev = st->st_dev; 2176 ost->st_ino = st->st_ino; 2177 ost->st_mode = st->st_mode; 2178 ost->st_nlink = st->st_nlink; 2179 ost->st_uid = st->st_uid; 2180 ost->st_gid = st->st_gid; 2181 ost->st_rdev = st->st_rdev; 2182 ost->st_size = MIN(st->st_size, INT32_MAX); 2183 ost->st_atim = st->st_atim; 2184 ost->st_mtim = st->st_mtim; 2185 ost->st_ctim = st->st_ctim; 2186 ost->st_blksize = st->st_blksize; 2187 ost->st_blocks = st->st_blocks; 2188 ost->st_flags = st->st_flags; 2189 ost->st_gen = st->st_gen; 2190 } 2191 #endif /* COMPAT_43 */ 2192 2193 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11) 2194 int ino64_trunc_error; 2195 SYSCTL_INT(_vfs, OID_AUTO, ino64_trunc_error, CTLFLAG_RW, 2196 &ino64_trunc_error, 0, 2197 "Error on truncation of device, file or inode number, or link count"); 2198 2199 int 2200 freebsd11_cvtstat(struct stat *st, struct freebsd11_stat *ost) 2201 { 2202 2203 ost->st_dev = st->st_dev; 2204 if (ost->st_dev != st->st_dev) { 2205 switch (ino64_trunc_error) { 2206 default: 2207 /* 2208 * Since dev_t is almost raw, don't clamp to the 2209 * maximum for case 2, but ignore the error. 2210 */ 2211 break; 2212 case 1: 2213 return (EOVERFLOW); 2214 } 2215 } 2216 ost->st_ino = st->st_ino; 2217 if (ost->st_ino != st->st_ino) { 2218 switch (ino64_trunc_error) { 2219 default: 2220 case 0: 2221 break; 2222 case 1: 2223 return (EOVERFLOW); 2224 case 2: 2225 ost->st_ino = UINT32_MAX; 2226 break; 2227 } 2228 } 2229 ost->st_mode = st->st_mode; 2230 ost->st_nlink = st->st_nlink; 2231 if (ost->st_nlink != st->st_nlink) { 2232 switch (ino64_trunc_error) { 2233 default: 2234 case 0: 2235 break; 2236 case 1: 2237 return (EOVERFLOW); 2238 case 2: 2239 ost->st_nlink = UINT16_MAX; 2240 break; 2241 } 2242 } 2243 ost->st_uid = st->st_uid; 2244 ost->st_gid = st->st_gid; 2245 ost->st_rdev = st->st_rdev; 2246 if (ost->st_rdev != st->st_rdev) { 2247 switch (ino64_trunc_error) { 2248 default: 2249 break; 2250 case 1: 2251 return (EOVERFLOW); 2252 } 2253 } 2254 ost->st_atim = st->st_atim; 2255 ost->st_mtim = st->st_mtim; 2256 ost->st_ctim = st->st_ctim; 2257 ost->st_size = st->st_size; 2258 ost->st_blocks = st->st_blocks; 2259 ost->st_blksize = st->st_blksize; 2260 ost->st_flags = st->st_flags; 2261 ost->st_gen = st->st_gen; 2262 ost->st_lspare = 0; 2263 ost->st_birthtim = st->st_birthtim; 2264 bzero((char *)&ost->st_birthtim + sizeof(ost->st_birthtim), 2265 sizeof(*ost) - offsetof(struct freebsd11_stat, 2266 st_birthtim) - sizeof(ost->st_birthtim)); 2267 return (0); 2268 } 2269 2270 int 2271 freebsd11_stat(struct thread *td, struct freebsd11_stat_args* uap) 2272 { 2273 struct stat sb; 2274 struct freebsd11_stat osb; 2275 int error; 2276 2277 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE, 2278 &sb, NULL); 2279 if (error != 0) 2280 return (error); 2281 error = freebsd11_cvtstat(&sb, &osb); 2282 if (error == 0) 2283 error = copyout(&osb, uap->ub, sizeof(osb)); 2284 return (error); 2285 } 2286 2287 int 2288 freebsd11_lstat(struct thread *td, struct freebsd11_lstat_args* uap) 2289 { 2290 struct stat sb; 2291 struct freebsd11_stat osb; 2292 int error; 2293 2294 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path, 2295 UIO_USERSPACE, &sb, NULL); 2296 if (error != 0) 2297 return (error); 2298 error = freebsd11_cvtstat(&sb, &osb); 2299 if (error == 0) 2300 error = copyout(&osb, uap->ub, sizeof(osb)); 2301 return (error); 2302 } 2303 2304 int 2305 freebsd11_fhstat(struct thread *td, struct freebsd11_fhstat_args* uap) 2306 { 2307 struct fhandle fh; 2308 struct stat sb; 2309 struct freebsd11_stat osb; 2310 int error; 2311 2312 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t)); 2313 if (error != 0) 2314 return (error); 2315 error = kern_fhstat(td, fh, &sb); 2316 if (error != 0) 2317 return (error); 2318 error = freebsd11_cvtstat(&sb, &osb); 2319 if (error == 0) 2320 error = copyout(&osb, uap->sb, sizeof(osb)); 2321 return (error); 2322 } 2323 2324 int 2325 freebsd11_fstatat(struct thread *td, struct freebsd11_fstatat_args* uap) 2326 { 2327 struct stat sb; 2328 struct freebsd11_stat osb; 2329 int error; 2330 2331 error = kern_statat(td, uap->flag, uap->fd, uap->path, 2332 UIO_USERSPACE, &sb, NULL); 2333 if (error != 0) 2334 return (error); 2335 error = freebsd11_cvtstat(&sb, &osb); 2336 if (error == 0) 2337 error = copyout(&osb, uap->buf, sizeof(osb)); 2338 return (error); 2339 } 2340 #endif /* COMPAT_FREEBSD11 */ 2341 2342 /* 2343 * Get file status 2344 */ 2345 #ifndef _SYS_SYSPROTO_H_ 2346 struct fstatat_args { 2347 int fd; 2348 char *path; 2349 struct stat *buf; 2350 int flag; 2351 } 2352 #endif 2353 int 2354 sys_fstatat(struct thread *td, struct fstatat_args *uap) 2355 { 2356 struct stat sb; 2357 int error; 2358 2359 error = kern_statat(td, uap->flag, uap->fd, uap->path, 2360 UIO_USERSPACE, &sb, NULL); 2361 if (error == 0) 2362 error = copyout(&sb, uap->buf, sizeof (sb)); 2363 return (error); 2364 } 2365 2366 int 2367 kern_statat(struct thread *td, int flag, int fd, const char *path, 2368 enum uio_seg pathseg, struct stat *sbp, 2369 void (*hook)(struct vnode *vp, struct stat *sbp)) 2370 { 2371 struct nameidata nd; 2372 int error; 2373 2374 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH)) != 0) 2375 return (EINVAL); 2376 2377 NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) != 0 ? 2378 NOFOLLOW : FOLLOW) | ((flag & AT_BENEATH) != 0 ? BENEATH : 0) | 2379 LOCKSHARED | LOCKLEAF | AUDITVNODE1, pathseg, path, fd, 2380 &cap_fstat_rights, td); 2381 2382 if ((error = namei(&nd)) != 0) 2383 return (error); 2384 error = vn_stat(nd.ni_vp, sbp, td->td_ucred, NOCRED, td); 2385 if (error == 0) { 2386 SDT_PROBE2(vfs, , stat, mode, path, sbp->st_mode); 2387 if (S_ISREG(sbp->st_mode)) 2388 SDT_PROBE2(vfs, , stat, reg, path, pathseg); 2389 if (__predict_false(hook != NULL)) 2390 hook(nd.ni_vp, sbp); 2391 } 2392 NDFREE(&nd, NDF_ONLY_PNBUF); 2393 vput(nd.ni_vp); 2394 #ifdef __STAT_TIME_T_EXT 2395 sbp->st_atim_ext = 0; 2396 sbp->st_mtim_ext = 0; 2397 sbp->st_ctim_ext = 0; 2398 sbp->st_btim_ext = 0; 2399 #endif 2400 #ifdef KTRACE 2401 if (KTRPOINT(td, KTR_STRUCT)) 2402 ktrstat_error(sbp, error); 2403 #endif 2404 return (error); 2405 } 2406 2407 #if defined(COMPAT_FREEBSD11) 2408 /* 2409 * Implementation of the NetBSD [l]stat() functions. 2410 */ 2411 void 2412 freebsd11_cvtnstat(struct stat *sb, struct nstat *nsb) 2413 { 2414 2415 bzero(nsb, sizeof(*nsb)); 2416 nsb->st_dev = sb->st_dev; 2417 nsb->st_ino = sb->st_ino; 2418 nsb->st_mode = sb->st_mode; 2419 nsb->st_nlink = sb->st_nlink; 2420 nsb->st_uid = sb->st_uid; 2421 nsb->st_gid = sb->st_gid; 2422 nsb->st_rdev = sb->st_rdev; 2423 nsb->st_atim = sb->st_atim; 2424 nsb->st_mtim = sb->st_mtim; 2425 nsb->st_ctim = sb->st_ctim; 2426 nsb->st_size = sb->st_size; 2427 nsb->st_blocks = sb->st_blocks; 2428 nsb->st_blksize = sb->st_blksize; 2429 nsb->st_flags = sb->st_flags; 2430 nsb->st_gen = sb->st_gen; 2431 nsb->st_birthtim = sb->st_birthtim; 2432 } 2433 2434 #ifndef _SYS_SYSPROTO_H_ 2435 struct freebsd11_nstat_args { 2436 char *path; 2437 struct nstat *ub; 2438 }; 2439 #endif 2440 int 2441 freebsd11_nstat(struct thread *td, struct freebsd11_nstat_args *uap) 2442 { 2443 struct stat sb; 2444 struct nstat nsb; 2445 int error; 2446 2447 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE, 2448 &sb, NULL); 2449 if (error != 0) 2450 return (error); 2451 freebsd11_cvtnstat(&sb, &nsb); 2452 return (copyout(&nsb, uap->ub, sizeof (nsb))); 2453 } 2454 2455 /* 2456 * NetBSD lstat. Get file status; this version does not follow links. 2457 */ 2458 #ifndef _SYS_SYSPROTO_H_ 2459 struct freebsd11_nlstat_args { 2460 char *path; 2461 struct nstat *ub; 2462 }; 2463 #endif 2464 int 2465 freebsd11_nlstat(struct thread *td, struct freebsd11_nlstat_args *uap) 2466 { 2467 struct stat sb; 2468 struct nstat nsb; 2469 int error; 2470 2471 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path, 2472 UIO_USERSPACE, &sb, NULL); 2473 if (error != 0) 2474 return (error); 2475 freebsd11_cvtnstat(&sb, &nsb); 2476 return (copyout(&nsb, uap->ub, sizeof (nsb))); 2477 } 2478 #endif /* COMPAT_FREEBSD11 */ 2479 2480 /* 2481 * Get configurable pathname variables. 2482 */ 2483 #ifndef _SYS_SYSPROTO_H_ 2484 struct pathconf_args { 2485 char *path; 2486 int name; 2487 }; 2488 #endif 2489 int 2490 sys_pathconf(struct thread *td, struct pathconf_args *uap) 2491 { 2492 long value; 2493 int error; 2494 2495 error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW, 2496 &value); 2497 if (error == 0) 2498 td->td_retval[0] = value; 2499 return (error); 2500 } 2501 2502 #ifndef _SYS_SYSPROTO_H_ 2503 struct lpathconf_args { 2504 char *path; 2505 int name; 2506 }; 2507 #endif 2508 int 2509 sys_lpathconf(struct thread *td, struct lpathconf_args *uap) 2510 { 2511 long value; 2512 int error; 2513 2514 error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, 2515 NOFOLLOW, &value); 2516 if (error == 0) 2517 td->td_retval[0] = value; 2518 return (error); 2519 } 2520 2521 int 2522 kern_pathconf(struct thread *td, const char *path, enum uio_seg pathseg, 2523 int name, u_long flags, long *valuep) 2524 { 2525 struct nameidata nd; 2526 int error; 2527 2528 NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags, 2529 pathseg, path, td); 2530 if ((error = namei(&nd)) != 0) 2531 return (error); 2532 NDFREE(&nd, NDF_ONLY_PNBUF); 2533 2534 error = VOP_PATHCONF(nd.ni_vp, name, valuep); 2535 vput(nd.ni_vp); 2536 return (error); 2537 } 2538 2539 /* 2540 * Return target name of a symbolic link. 2541 */ 2542 #ifndef _SYS_SYSPROTO_H_ 2543 struct readlink_args { 2544 char *path; 2545 char *buf; 2546 size_t count; 2547 }; 2548 #endif 2549 int 2550 sys_readlink(struct thread *td, struct readlink_args *uap) 2551 { 2552 2553 return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2554 uap->buf, UIO_USERSPACE, uap->count)); 2555 } 2556 #ifndef _SYS_SYSPROTO_H_ 2557 struct readlinkat_args { 2558 int fd; 2559 char *path; 2560 char *buf; 2561 size_t bufsize; 2562 }; 2563 #endif 2564 int 2565 sys_readlinkat(struct thread *td, struct readlinkat_args *uap) 2566 { 2567 2568 return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE, 2569 uap->buf, UIO_USERSPACE, uap->bufsize)); 2570 } 2571 2572 int 2573 kern_readlinkat(struct thread *td, int fd, const char *path, 2574 enum uio_seg pathseg, char *buf, enum uio_seg bufseg, size_t count) 2575 { 2576 struct vnode *vp; 2577 struct nameidata nd; 2578 int error; 2579 2580 if (count > IOSIZE_MAX) 2581 return (EINVAL); 2582 2583 NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1, 2584 pathseg, path, fd, td); 2585 2586 if ((error = namei(&nd)) != 0) 2587 return (error); 2588 NDFREE(&nd, NDF_ONLY_PNBUF); 2589 vp = nd.ni_vp; 2590 2591 error = kern_readlink_vp(vp, buf, bufseg, count, td); 2592 vput(vp); 2593 2594 return (error); 2595 } 2596 2597 /* 2598 * Helper function to readlink from a vnode 2599 */ 2600 static int 2601 kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg, size_t count, 2602 struct thread *td) 2603 { 2604 struct iovec aiov; 2605 struct uio auio; 2606 int error; 2607 2608 ASSERT_VOP_LOCKED(vp, "kern_readlink_vp(): vp not locked"); 2609 #ifdef MAC 2610 error = mac_vnode_check_readlink(td->td_ucred, vp); 2611 if (error != 0) 2612 return (error); 2613 #endif 2614 if (vp->v_type != VLNK && (vp->v_vflag & VV_READLINK) == 0) 2615 return (EINVAL); 2616 2617 aiov.iov_base = buf; 2618 aiov.iov_len = count; 2619 auio.uio_iov = &aiov; 2620 auio.uio_iovcnt = 1; 2621 auio.uio_offset = 0; 2622 auio.uio_rw = UIO_READ; 2623 auio.uio_segflg = bufseg; 2624 auio.uio_td = td; 2625 auio.uio_resid = count; 2626 error = VOP_READLINK(vp, &auio, td->td_ucred); 2627 td->td_retval[0] = count - auio.uio_resid; 2628 return (error); 2629 } 2630 2631 /* 2632 * Common implementation code for chflags() and fchflags(). 2633 */ 2634 static int 2635 setfflags(struct thread *td, struct vnode *vp, u_long flags) 2636 { 2637 struct mount *mp; 2638 struct vattr vattr; 2639 int error; 2640 2641 /* We can't support the value matching VNOVAL. */ 2642 if (flags == VNOVAL) 2643 return (EOPNOTSUPP); 2644 2645 /* 2646 * Prevent non-root users from setting flags on devices. When 2647 * a device is reused, users can retain ownership of the device 2648 * if they are allowed to set flags and programs assume that 2649 * chown can't fail when done as root. 2650 */ 2651 if (vp->v_type == VCHR || vp->v_type == VBLK) { 2652 error = priv_check(td, PRIV_VFS_CHFLAGS_DEV); 2653 if (error != 0) 2654 return (error); 2655 } 2656 2657 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) 2658 return (error); 2659 VATTR_NULL(&vattr); 2660 vattr.va_flags = flags; 2661 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2662 #ifdef MAC 2663 error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags); 2664 if (error == 0) 2665 #endif 2666 error = VOP_SETATTR(vp, &vattr, td->td_ucred); 2667 VOP_UNLOCK(vp); 2668 vn_finished_write(mp); 2669 return (error); 2670 } 2671 2672 /* 2673 * Change flags of a file given a path name. 2674 */ 2675 #ifndef _SYS_SYSPROTO_H_ 2676 struct chflags_args { 2677 const char *path; 2678 u_long flags; 2679 }; 2680 #endif 2681 int 2682 sys_chflags(struct thread *td, struct chflags_args *uap) 2683 { 2684 2685 return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2686 uap->flags, 0)); 2687 } 2688 2689 #ifndef _SYS_SYSPROTO_H_ 2690 struct chflagsat_args { 2691 int fd; 2692 const char *path; 2693 u_long flags; 2694 int atflag; 2695 } 2696 #endif 2697 int 2698 sys_chflagsat(struct thread *td, struct chflagsat_args *uap) 2699 { 2700 2701 if ((uap->atflag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH)) != 0) 2702 return (EINVAL); 2703 2704 return (kern_chflagsat(td, uap->fd, uap->path, UIO_USERSPACE, 2705 uap->flags, uap->atflag)); 2706 } 2707 2708 /* 2709 * Same as chflags() but doesn't follow symlinks. 2710 */ 2711 #ifndef _SYS_SYSPROTO_H_ 2712 struct lchflags_args { 2713 const char *path; 2714 u_long flags; 2715 }; 2716 #endif 2717 int 2718 sys_lchflags(struct thread *td, struct lchflags_args *uap) 2719 { 2720 2721 return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2722 uap->flags, AT_SYMLINK_NOFOLLOW)); 2723 } 2724 2725 static int 2726 kern_chflagsat(struct thread *td, int fd, const char *path, 2727 enum uio_seg pathseg, u_long flags, int atflag) 2728 { 2729 struct nameidata nd; 2730 int error, follow; 2731 2732 AUDIT_ARG_FFLAGS(flags); 2733 follow = (atflag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW; 2734 follow |= (atflag & AT_BENEATH) != 0 ? BENEATH : 0; 2735 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd, 2736 &cap_fchflags_rights, td); 2737 if ((error = namei(&nd)) != 0) 2738 return (error); 2739 NDFREE(&nd, NDF_ONLY_PNBUF); 2740 error = setfflags(td, nd.ni_vp, flags); 2741 vrele(nd.ni_vp); 2742 return (error); 2743 } 2744 2745 /* 2746 * Change flags of a file given a file descriptor. 2747 */ 2748 #ifndef _SYS_SYSPROTO_H_ 2749 struct fchflags_args { 2750 int fd; 2751 u_long flags; 2752 }; 2753 #endif 2754 int 2755 sys_fchflags(struct thread *td, struct fchflags_args *uap) 2756 { 2757 struct file *fp; 2758 int error; 2759 2760 AUDIT_ARG_FD(uap->fd); 2761 AUDIT_ARG_FFLAGS(uap->flags); 2762 error = getvnode(td, uap->fd, &cap_fchflags_rights, 2763 &fp); 2764 if (error != 0) 2765 return (error); 2766 #ifdef AUDIT 2767 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY); 2768 AUDIT_ARG_VNODE1(fp->f_vnode); 2769 VOP_UNLOCK(fp->f_vnode); 2770 #endif 2771 error = setfflags(td, fp->f_vnode, uap->flags); 2772 fdrop(fp, td); 2773 return (error); 2774 } 2775 2776 /* 2777 * Common implementation code for chmod(), lchmod() and fchmod(). 2778 */ 2779 int 2780 setfmode(struct thread *td, struct ucred *cred, struct vnode *vp, int mode) 2781 { 2782 struct mount *mp; 2783 struct vattr vattr; 2784 int error; 2785 2786 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) 2787 return (error); 2788 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2789 VATTR_NULL(&vattr); 2790 vattr.va_mode = mode & ALLPERMS; 2791 #ifdef MAC 2792 error = mac_vnode_check_setmode(cred, vp, vattr.va_mode); 2793 if (error == 0) 2794 #endif 2795 error = VOP_SETATTR(vp, &vattr, cred); 2796 VOP_UNLOCK(vp); 2797 vn_finished_write(mp); 2798 return (error); 2799 } 2800 2801 /* 2802 * Change mode of a file given path name. 2803 */ 2804 #ifndef _SYS_SYSPROTO_H_ 2805 struct chmod_args { 2806 char *path; 2807 int mode; 2808 }; 2809 #endif 2810 int 2811 sys_chmod(struct thread *td, struct chmod_args *uap) 2812 { 2813 2814 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2815 uap->mode, 0)); 2816 } 2817 2818 #ifndef _SYS_SYSPROTO_H_ 2819 struct fchmodat_args { 2820 int dirfd; 2821 char *path; 2822 mode_t mode; 2823 int flag; 2824 } 2825 #endif 2826 int 2827 sys_fchmodat(struct thread *td, struct fchmodat_args *uap) 2828 { 2829 2830 if ((uap->flag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH)) != 0) 2831 return (EINVAL); 2832 2833 return (kern_fchmodat(td, uap->fd, uap->path, UIO_USERSPACE, 2834 uap->mode, uap->flag)); 2835 } 2836 2837 /* 2838 * Change mode of a file given path name (don't follow links.) 2839 */ 2840 #ifndef _SYS_SYSPROTO_H_ 2841 struct lchmod_args { 2842 char *path; 2843 int mode; 2844 }; 2845 #endif 2846 int 2847 sys_lchmod(struct thread *td, struct lchmod_args *uap) 2848 { 2849 2850 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2851 uap->mode, AT_SYMLINK_NOFOLLOW)); 2852 } 2853 2854 int 2855 kern_fchmodat(struct thread *td, int fd, const char *path, 2856 enum uio_seg pathseg, mode_t mode, int flag) 2857 { 2858 struct nameidata nd; 2859 int error, follow; 2860 2861 AUDIT_ARG_MODE(mode); 2862 follow = (flag & AT_SYMLINK_NOFOLLOW) != 0 ? NOFOLLOW : FOLLOW; 2863 follow |= (flag & AT_BENEATH) != 0 ? BENEATH : 0; 2864 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd, 2865 &cap_fchmod_rights, td); 2866 if ((error = namei(&nd)) != 0) 2867 return (error); 2868 NDFREE(&nd, NDF_ONLY_PNBUF); 2869 error = setfmode(td, td->td_ucred, nd.ni_vp, mode); 2870 vrele(nd.ni_vp); 2871 return (error); 2872 } 2873 2874 /* 2875 * Change mode of a file given a file descriptor. 2876 */ 2877 #ifndef _SYS_SYSPROTO_H_ 2878 struct fchmod_args { 2879 int fd; 2880 int mode; 2881 }; 2882 #endif 2883 int 2884 sys_fchmod(struct thread *td, struct fchmod_args *uap) 2885 { 2886 struct file *fp; 2887 int error; 2888 2889 AUDIT_ARG_FD(uap->fd); 2890 AUDIT_ARG_MODE(uap->mode); 2891 2892 error = fget(td, uap->fd, &cap_fchmod_rights, &fp); 2893 if (error != 0) 2894 return (error); 2895 error = fo_chmod(fp, uap->mode, td->td_ucred, td); 2896 fdrop(fp, td); 2897 return (error); 2898 } 2899 2900 /* 2901 * Common implementation for chown(), lchown(), and fchown() 2902 */ 2903 int 2904 setfown(struct thread *td, struct ucred *cred, struct vnode *vp, uid_t uid, 2905 gid_t gid) 2906 { 2907 struct mount *mp; 2908 struct vattr vattr; 2909 int error; 2910 2911 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) 2912 return (error); 2913 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2914 VATTR_NULL(&vattr); 2915 vattr.va_uid = uid; 2916 vattr.va_gid = gid; 2917 #ifdef MAC 2918 error = mac_vnode_check_setowner(cred, vp, vattr.va_uid, 2919 vattr.va_gid); 2920 if (error == 0) 2921 #endif 2922 error = VOP_SETATTR(vp, &vattr, cred); 2923 VOP_UNLOCK(vp); 2924 vn_finished_write(mp); 2925 return (error); 2926 } 2927 2928 /* 2929 * Set ownership given a path name. 2930 */ 2931 #ifndef _SYS_SYSPROTO_H_ 2932 struct chown_args { 2933 char *path; 2934 int uid; 2935 int gid; 2936 }; 2937 #endif 2938 int 2939 sys_chown(struct thread *td, struct chown_args *uap) 2940 { 2941 2942 return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid, 2943 uap->gid, 0)); 2944 } 2945 2946 #ifndef _SYS_SYSPROTO_H_ 2947 struct fchownat_args { 2948 int fd; 2949 const char * path; 2950 uid_t uid; 2951 gid_t gid; 2952 int flag; 2953 }; 2954 #endif 2955 int 2956 sys_fchownat(struct thread *td, struct fchownat_args *uap) 2957 { 2958 2959 if ((uap->flag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH)) != 0) 2960 return (EINVAL); 2961 2962 return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid, 2963 uap->gid, uap->flag)); 2964 } 2965 2966 int 2967 kern_fchownat(struct thread *td, int fd, const char *path, 2968 enum uio_seg pathseg, int uid, int gid, int flag) 2969 { 2970 struct nameidata nd; 2971 int error, follow; 2972 2973 AUDIT_ARG_OWNER(uid, gid); 2974 follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW; 2975 follow |= (flag & AT_BENEATH) != 0 ? BENEATH : 0; 2976 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd, 2977 &cap_fchown_rights, td); 2978 2979 if ((error = namei(&nd)) != 0) 2980 return (error); 2981 NDFREE(&nd, NDF_ONLY_PNBUF); 2982 error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid); 2983 vrele(nd.ni_vp); 2984 return (error); 2985 } 2986 2987 /* 2988 * Set ownership given a path name, do not cross symlinks. 2989 */ 2990 #ifndef _SYS_SYSPROTO_H_ 2991 struct lchown_args { 2992 char *path; 2993 int uid; 2994 int gid; 2995 }; 2996 #endif 2997 int 2998 sys_lchown(struct thread *td, struct lchown_args *uap) 2999 { 3000 3001 return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 3002 uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW)); 3003 } 3004 3005 /* 3006 * Set ownership given a file descriptor. 3007 */ 3008 #ifndef _SYS_SYSPROTO_H_ 3009 struct fchown_args { 3010 int fd; 3011 int uid; 3012 int gid; 3013 }; 3014 #endif 3015 int 3016 sys_fchown(struct thread *td, struct fchown_args *uap) 3017 { 3018 struct file *fp; 3019 int error; 3020 3021 AUDIT_ARG_FD(uap->fd); 3022 AUDIT_ARG_OWNER(uap->uid, uap->gid); 3023 error = fget(td, uap->fd, &cap_fchown_rights, &fp); 3024 if (error != 0) 3025 return (error); 3026 error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td); 3027 fdrop(fp, td); 3028 return (error); 3029 } 3030 3031 /* 3032 * Common implementation code for utimes(), lutimes(), and futimes(). 3033 */ 3034 static int 3035 getutimes(const struct timeval *usrtvp, enum uio_seg tvpseg, 3036 struct timespec *tsp) 3037 { 3038 struct timeval tv[2]; 3039 const struct timeval *tvp; 3040 int error; 3041 3042 if (usrtvp == NULL) { 3043 vfs_timestamp(&tsp[0]); 3044 tsp[1] = tsp[0]; 3045 } else { 3046 if (tvpseg == UIO_SYSSPACE) { 3047 tvp = usrtvp; 3048 } else { 3049 if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0) 3050 return (error); 3051 tvp = tv; 3052 } 3053 3054 if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 || 3055 tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000) 3056 return (EINVAL); 3057 TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]); 3058 TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]); 3059 } 3060 return (0); 3061 } 3062 3063 /* 3064 * Common implementation code for futimens(), utimensat(). 3065 */ 3066 #define UTIMENS_NULL 0x1 3067 #define UTIMENS_EXIT 0x2 3068 static int 3069 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg, 3070 struct timespec *tsp, int *retflags) 3071 { 3072 struct timespec tsnow; 3073 int error; 3074 3075 vfs_timestamp(&tsnow); 3076 *retflags = 0; 3077 if (usrtsp == NULL) { 3078 tsp[0] = tsnow; 3079 tsp[1] = tsnow; 3080 *retflags |= UTIMENS_NULL; 3081 return (0); 3082 } 3083 if (tspseg == UIO_SYSSPACE) { 3084 tsp[0] = usrtsp[0]; 3085 tsp[1] = usrtsp[1]; 3086 } else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0) 3087 return (error); 3088 if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT) 3089 *retflags |= UTIMENS_EXIT; 3090 if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW) 3091 *retflags |= UTIMENS_NULL; 3092 if (tsp[0].tv_nsec == UTIME_OMIT) 3093 tsp[0].tv_sec = VNOVAL; 3094 else if (tsp[0].tv_nsec == UTIME_NOW) 3095 tsp[0] = tsnow; 3096 else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L) 3097 return (EINVAL); 3098 if (tsp[1].tv_nsec == UTIME_OMIT) 3099 tsp[1].tv_sec = VNOVAL; 3100 else if (tsp[1].tv_nsec == UTIME_NOW) 3101 tsp[1] = tsnow; 3102 else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L) 3103 return (EINVAL); 3104 3105 return (0); 3106 } 3107 3108 /* 3109 * Common implementation code for utimes(), lutimes(), futimes(), futimens(), 3110 * and utimensat(). 3111 */ 3112 static int 3113 setutimes(struct thread *td, struct vnode *vp, const struct timespec *ts, 3114 int numtimes, int nullflag) 3115 { 3116 struct mount *mp; 3117 struct vattr vattr; 3118 int error, setbirthtime; 3119 3120 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) 3121 return (error); 3122 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3123 setbirthtime = 0; 3124 if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) && 3125 timespeccmp(&ts[1], &vattr.va_birthtime, < )) 3126 setbirthtime = 1; 3127 VATTR_NULL(&vattr); 3128 vattr.va_atime = ts[0]; 3129 vattr.va_mtime = ts[1]; 3130 if (setbirthtime) 3131 vattr.va_birthtime = ts[1]; 3132 if (numtimes > 2) 3133 vattr.va_birthtime = ts[2]; 3134 if (nullflag) 3135 vattr.va_vaflags |= VA_UTIMES_NULL; 3136 #ifdef MAC 3137 error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime, 3138 vattr.va_mtime); 3139 #endif 3140 if (error == 0) 3141 error = VOP_SETATTR(vp, &vattr, td->td_ucred); 3142 VOP_UNLOCK(vp); 3143 vn_finished_write(mp); 3144 return (error); 3145 } 3146 3147 /* 3148 * Set the access and modification times of a file. 3149 */ 3150 #ifndef _SYS_SYSPROTO_H_ 3151 struct utimes_args { 3152 char *path; 3153 struct timeval *tptr; 3154 }; 3155 #endif 3156 int 3157 sys_utimes(struct thread *td, struct utimes_args *uap) 3158 { 3159 3160 return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 3161 uap->tptr, UIO_USERSPACE)); 3162 } 3163 3164 #ifndef _SYS_SYSPROTO_H_ 3165 struct futimesat_args { 3166 int fd; 3167 const char * path; 3168 const struct timeval * times; 3169 }; 3170 #endif 3171 int 3172 sys_futimesat(struct thread *td, struct futimesat_args *uap) 3173 { 3174 3175 return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE, 3176 uap->times, UIO_USERSPACE)); 3177 } 3178 3179 int 3180 kern_utimesat(struct thread *td, int fd, const char *path, 3181 enum uio_seg pathseg, struct timeval *tptr, enum uio_seg tptrseg) 3182 { 3183 struct nameidata nd; 3184 struct timespec ts[2]; 3185 int error; 3186 3187 if ((error = getutimes(tptr, tptrseg, ts)) != 0) 3188 return (error); 3189 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd, 3190 &cap_futimes_rights, td); 3191 3192 if ((error = namei(&nd)) != 0) 3193 return (error); 3194 NDFREE(&nd, NDF_ONLY_PNBUF); 3195 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL); 3196 vrele(nd.ni_vp); 3197 return (error); 3198 } 3199 3200 /* 3201 * Set the access and modification times of a file. 3202 */ 3203 #ifndef _SYS_SYSPROTO_H_ 3204 struct lutimes_args { 3205 char *path; 3206 struct timeval *tptr; 3207 }; 3208 #endif 3209 int 3210 sys_lutimes(struct thread *td, struct lutimes_args *uap) 3211 { 3212 3213 return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr, 3214 UIO_USERSPACE)); 3215 } 3216 3217 int 3218 kern_lutimes(struct thread *td, const char *path, enum uio_seg pathseg, 3219 struct timeval *tptr, enum uio_seg tptrseg) 3220 { 3221 struct timespec ts[2]; 3222 struct nameidata nd; 3223 int error; 3224 3225 if ((error = getutimes(tptr, tptrseg, ts)) != 0) 3226 return (error); 3227 NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td); 3228 if ((error = namei(&nd)) != 0) 3229 return (error); 3230 NDFREE(&nd, NDF_ONLY_PNBUF); 3231 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL); 3232 vrele(nd.ni_vp); 3233 return (error); 3234 } 3235 3236 /* 3237 * Set the access and modification times of a file. 3238 */ 3239 #ifndef _SYS_SYSPROTO_H_ 3240 struct futimes_args { 3241 int fd; 3242 struct timeval *tptr; 3243 }; 3244 #endif 3245 int 3246 sys_futimes(struct thread *td, struct futimes_args *uap) 3247 { 3248 3249 return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE)); 3250 } 3251 3252 int 3253 kern_futimes(struct thread *td, int fd, struct timeval *tptr, 3254 enum uio_seg tptrseg) 3255 { 3256 struct timespec ts[2]; 3257 struct file *fp; 3258 int error; 3259 3260 AUDIT_ARG_FD(fd); 3261 error = getutimes(tptr, tptrseg, ts); 3262 if (error != 0) 3263 return (error); 3264 error = getvnode(td, fd, &cap_futimes_rights, &fp); 3265 if (error != 0) 3266 return (error); 3267 #ifdef AUDIT 3268 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY); 3269 AUDIT_ARG_VNODE1(fp->f_vnode); 3270 VOP_UNLOCK(fp->f_vnode); 3271 #endif 3272 error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL); 3273 fdrop(fp, td); 3274 return (error); 3275 } 3276 3277 int 3278 sys_futimens(struct thread *td, struct futimens_args *uap) 3279 { 3280 3281 return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE)); 3282 } 3283 3284 int 3285 kern_futimens(struct thread *td, int fd, struct timespec *tptr, 3286 enum uio_seg tptrseg) 3287 { 3288 struct timespec ts[2]; 3289 struct file *fp; 3290 int error, flags; 3291 3292 AUDIT_ARG_FD(fd); 3293 error = getutimens(tptr, tptrseg, ts, &flags); 3294 if (error != 0) 3295 return (error); 3296 if (flags & UTIMENS_EXIT) 3297 return (0); 3298 error = getvnode(td, fd, &cap_futimes_rights, &fp); 3299 if (error != 0) 3300 return (error); 3301 #ifdef AUDIT 3302 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY); 3303 AUDIT_ARG_VNODE1(fp->f_vnode); 3304 VOP_UNLOCK(fp->f_vnode); 3305 #endif 3306 error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL); 3307 fdrop(fp, td); 3308 return (error); 3309 } 3310 3311 int 3312 sys_utimensat(struct thread *td, struct utimensat_args *uap) 3313 { 3314 3315 return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE, 3316 uap->times, UIO_USERSPACE, uap->flag)); 3317 } 3318 3319 int 3320 kern_utimensat(struct thread *td, int fd, const char *path, 3321 enum uio_seg pathseg, struct timespec *tptr, enum uio_seg tptrseg, 3322 int flag) 3323 { 3324 struct nameidata nd; 3325 struct timespec ts[2]; 3326 int error, flags; 3327 3328 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH)) != 0) 3329 return (EINVAL); 3330 3331 if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0) 3332 return (error); 3333 NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : 3334 FOLLOW) | ((flag & AT_BENEATH) != 0 ? BENEATH : 0) | AUDITVNODE1, 3335 pathseg, path, fd, &cap_futimes_rights, td); 3336 if ((error = namei(&nd)) != 0) 3337 return (error); 3338 /* 3339 * We are allowed to call namei() regardless of 2xUTIME_OMIT. 3340 * POSIX states: 3341 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected." 3342 * "Search permission is denied by a component of the path prefix." 3343 */ 3344 NDFREE(&nd, NDF_ONLY_PNBUF); 3345 if ((flags & UTIMENS_EXIT) == 0) 3346 error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL); 3347 vrele(nd.ni_vp); 3348 return (error); 3349 } 3350 3351 /* 3352 * Truncate a file given its path name. 3353 */ 3354 #ifndef _SYS_SYSPROTO_H_ 3355 struct truncate_args { 3356 char *path; 3357 int pad; 3358 off_t length; 3359 }; 3360 #endif 3361 int 3362 sys_truncate(struct thread *td, struct truncate_args *uap) 3363 { 3364 3365 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length)); 3366 } 3367 3368 int 3369 kern_truncate(struct thread *td, const char *path, enum uio_seg pathseg, 3370 off_t length) 3371 { 3372 struct mount *mp; 3373 struct vnode *vp; 3374 void *rl_cookie; 3375 struct vattr vattr; 3376 struct nameidata nd; 3377 int error; 3378 3379 if (length < 0) 3380 return(EINVAL); 3381 NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td); 3382 if ((error = namei(&nd)) != 0) 3383 return (error); 3384 vp = nd.ni_vp; 3385 rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX); 3386 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) { 3387 vn_rangelock_unlock(vp, rl_cookie); 3388 vrele(vp); 3389 return (error); 3390 } 3391 NDFREE(&nd, NDF_ONLY_PNBUF); 3392 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3393 if (vp->v_type == VDIR) 3394 error = EISDIR; 3395 #ifdef MAC 3396 else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) { 3397 } 3398 #endif 3399 else if ((error = vn_writechk(vp)) == 0 && 3400 (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) { 3401 VATTR_NULL(&vattr); 3402 vattr.va_size = length; 3403 error = VOP_SETATTR(vp, &vattr, td->td_ucred); 3404 } 3405 VOP_UNLOCK(vp); 3406 vn_finished_write(mp); 3407 vn_rangelock_unlock(vp, rl_cookie); 3408 vrele(vp); 3409 return (error); 3410 } 3411 3412 #if defined(COMPAT_43) 3413 /* 3414 * Truncate a file given its path name. 3415 */ 3416 #ifndef _SYS_SYSPROTO_H_ 3417 struct otruncate_args { 3418 char *path; 3419 long length; 3420 }; 3421 #endif 3422 int 3423 otruncate(struct thread *td, struct otruncate_args *uap) 3424 { 3425 3426 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length)); 3427 } 3428 #endif /* COMPAT_43 */ 3429 3430 #if defined(COMPAT_FREEBSD6) 3431 /* Versions with the pad argument */ 3432 int 3433 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap) 3434 { 3435 3436 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length)); 3437 } 3438 3439 int 3440 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap) 3441 { 3442 3443 return (kern_ftruncate(td, uap->fd, uap->length)); 3444 } 3445 #endif 3446 3447 int 3448 kern_fsync(struct thread *td, int fd, bool fullsync) 3449 { 3450 struct vnode *vp; 3451 struct mount *mp; 3452 struct file *fp; 3453 int error, lock_flags; 3454 3455 AUDIT_ARG_FD(fd); 3456 error = getvnode(td, fd, &cap_fsync_rights, &fp); 3457 if (error != 0) 3458 return (error); 3459 vp = fp->f_vnode; 3460 #if 0 3461 if (!fullsync) 3462 /* XXXKIB: compete outstanding aio writes */; 3463 #endif 3464 error = vn_start_write(vp, &mp, V_WAIT | PCATCH); 3465 if (error != 0) 3466 goto drop; 3467 if (MNT_SHARED_WRITES(mp) || 3468 ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) { 3469 lock_flags = LK_SHARED; 3470 } else { 3471 lock_flags = LK_EXCLUSIVE; 3472 } 3473 vn_lock(vp, lock_flags | LK_RETRY); 3474 AUDIT_ARG_VNODE1(vp); 3475 if (vp->v_object != NULL) { 3476 VM_OBJECT_WLOCK(vp->v_object); 3477 vm_object_page_clean(vp->v_object, 0, 0, 0); 3478 VM_OBJECT_WUNLOCK(vp->v_object); 3479 } 3480 error = fullsync ? VOP_FSYNC(vp, MNT_WAIT, td) : VOP_FDATASYNC(vp, td); 3481 VOP_UNLOCK(vp); 3482 vn_finished_write(mp); 3483 drop: 3484 fdrop(fp, td); 3485 return (error); 3486 } 3487 3488 /* 3489 * Sync an open file. 3490 */ 3491 #ifndef _SYS_SYSPROTO_H_ 3492 struct fsync_args { 3493 int fd; 3494 }; 3495 #endif 3496 int 3497 sys_fsync(struct thread *td, struct fsync_args *uap) 3498 { 3499 3500 return (kern_fsync(td, uap->fd, true)); 3501 } 3502 3503 int 3504 sys_fdatasync(struct thread *td, struct fdatasync_args *uap) 3505 { 3506 3507 return (kern_fsync(td, uap->fd, false)); 3508 } 3509 3510 /* 3511 * Rename files. Source and destination must either both be directories, or 3512 * both not be directories. If target is a directory, it must be empty. 3513 */ 3514 #ifndef _SYS_SYSPROTO_H_ 3515 struct rename_args { 3516 char *from; 3517 char *to; 3518 }; 3519 #endif 3520 int 3521 sys_rename(struct thread *td, struct rename_args *uap) 3522 { 3523 3524 return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD, 3525 uap->to, UIO_USERSPACE)); 3526 } 3527 3528 #ifndef _SYS_SYSPROTO_H_ 3529 struct renameat_args { 3530 int oldfd; 3531 char *old; 3532 int newfd; 3533 char *new; 3534 }; 3535 #endif 3536 int 3537 sys_renameat(struct thread *td, struct renameat_args *uap) 3538 { 3539 3540 return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new, 3541 UIO_USERSPACE)); 3542 } 3543 3544 int 3545 kern_renameat(struct thread *td, int oldfd, const char *old, int newfd, 3546 const char *new, enum uio_seg pathseg) 3547 { 3548 struct mount *mp = NULL; 3549 struct vnode *tvp, *fvp, *tdvp; 3550 struct nameidata fromnd, tond; 3551 int error; 3552 3553 again: 3554 bwillwrite(); 3555 #ifdef MAC 3556 NDINIT_ATRIGHTS(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART | 3557 AUDITVNODE1, pathseg, old, oldfd, 3558 &cap_renameat_source_rights, td); 3559 #else 3560 NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1, 3561 pathseg, old, oldfd, 3562 &cap_renameat_source_rights, td); 3563 #endif 3564 3565 if ((error = namei(&fromnd)) != 0) 3566 return (error); 3567 #ifdef MAC 3568 error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp, 3569 fromnd.ni_vp, &fromnd.ni_cnd); 3570 VOP_UNLOCK(fromnd.ni_dvp); 3571 if (fromnd.ni_dvp != fromnd.ni_vp) 3572 VOP_UNLOCK(fromnd.ni_vp); 3573 #endif 3574 fvp = fromnd.ni_vp; 3575 NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | 3576 SAVESTART | AUDITVNODE2, pathseg, new, newfd, 3577 &cap_renameat_target_rights, td); 3578 if (fromnd.ni_vp->v_type == VDIR) 3579 tond.ni_cnd.cn_flags |= WILLBEDIR; 3580 if ((error = namei(&tond)) != 0) { 3581 /* Translate error code for rename("dir1", "dir2/."). */ 3582 if (error == EISDIR && fvp->v_type == VDIR) 3583 error = EINVAL; 3584 NDFREE(&fromnd, NDF_ONLY_PNBUF); 3585 vrele(fromnd.ni_dvp); 3586 vrele(fvp); 3587 goto out1; 3588 } 3589 tdvp = tond.ni_dvp; 3590 tvp = tond.ni_vp; 3591 error = vn_start_write(fvp, &mp, V_NOWAIT); 3592 if (error != 0) { 3593 NDFREE(&fromnd, NDF_ONLY_PNBUF); 3594 NDFREE(&tond, NDF_ONLY_PNBUF); 3595 if (tvp != NULL) 3596 vput(tvp); 3597 if (tdvp == tvp) 3598 vrele(tdvp); 3599 else 3600 vput(tdvp); 3601 vrele(fromnd.ni_dvp); 3602 vrele(fvp); 3603 vrele(tond.ni_startdir); 3604 if (fromnd.ni_startdir != NULL) 3605 vrele(fromnd.ni_startdir); 3606 error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH); 3607 if (error != 0) 3608 return (error); 3609 goto again; 3610 } 3611 if (tvp != NULL) { 3612 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3613 error = ENOTDIR; 3614 goto out; 3615 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3616 error = EISDIR; 3617 goto out; 3618 } 3619 #ifdef CAPABILITIES 3620 if (newfd != AT_FDCWD && (tond.ni_resflags & NIRES_ABS) == 0) { 3621 /* 3622 * If the target already exists we require CAP_UNLINKAT 3623 * from 'newfd', when newfd was used for the lookup. 3624 */ 3625 error = cap_check(&tond.ni_filecaps.fc_rights, 3626 &cap_unlinkat_rights); 3627 if (error != 0) 3628 goto out; 3629 } 3630 #endif 3631 } 3632 if (fvp == tdvp) { 3633 error = EINVAL; 3634 goto out; 3635 } 3636 /* 3637 * If the source is the same as the destination (that is, if they 3638 * are links to the same vnode), then there is nothing to do. 3639 */ 3640 if (fvp == tvp) 3641 error = -1; 3642 #ifdef MAC 3643 else 3644 error = mac_vnode_check_rename_to(td->td_ucred, tdvp, 3645 tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd); 3646 #endif 3647 out: 3648 if (error == 0) { 3649 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3650 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3651 NDFREE(&fromnd, NDF_ONLY_PNBUF); 3652 NDFREE(&tond, NDF_ONLY_PNBUF); 3653 } else { 3654 NDFREE(&fromnd, NDF_ONLY_PNBUF); 3655 NDFREE(&tond, NDF_ONLY_PNBUF); 3656 if (tvp != NULL) 3657 vput(tvp); 3658 if (tdvp == tvp) 3659 vrele(tdvp); 3660 else 3661 vput(tdvp); 3662 vrele(fromnd.ni_dvp); 3663 vrele(fvp); 3664 } 3665 vrele(tond.ni_startdir); 3666 vn_finished_write(mp); 3667 out1: 3668 if (fromnd.ni_startdir) 3669 vrele(fromnd.ni_startdir); 3670 if (error == -1) 3671 return (0); 3672 return (error); 3673 } 3674 3675 /* 3676 * Make a directory file. 3677 */ 3678 #ifndef _SYS_SYSPROTO_H_ 3679 struct mkdir_args { 3680 char *path; 3681 int mode; 3682 }; 3683 #endif 3684 int 3685 sys_mkdir(struct thread *td, struct mkdir_args *uap) 3686 { 3687 3688 return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 3689 uap->mode)); 3690 } 3691 3692 #ifndef _SYS_SYSPROTO_H_ 3693 struct mkdirat_args { 3694 int fd; 3695 char *path; 3696 mode_t mode; 3697 }; 3698 #endif 3699 int 3700 sys_mkdirat(struct thread *td, struct mkdirat_args *uap) 3701 { 3702 3703 return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode)); 3704 } 3705 3706 int 3707 kern_mkdirat(struct thread *td, int fd, const char *path, enum uio_seg segflg, 3708 int mode) 3709 { 3710 struct mount *mp; 3711 struct vnode *vp; 3712 struct vattr vattr; 3713 struct nameidata nd; 3714 int error; 3715 3716 AUDIT_ARG_MODE(mode); 3717 restart: 3718 bwillwrite(); 3719 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 | 3720 NOCACHE, segflg, path, fd, &cap_mkdirat_rights, 3721 td); 3722 nd.ni_cnd.cn_flags |= WILLBEDIR; 3723 if ((error = namei(&nd)) != 0) 3724 return (error); 3725 vp = nd.ni_vp; 3726 if (vp != NULL) { 3727 NDFREE(&nd, NDF_ONLY_PNBUF); 3728 /* 3729 * XXX namei called with LOCKPARENT but not LOCKLEAF has 3730 * the strange behaviour of leaving the vnode unlocked 3731 * if the target is the same vnode as the parent. 3732 */ 3733 if (vp == nd.ni_dvp) 3734 vrele(nd.ni_dvp); 3735 else 3736 vput(nd.ni_dvp); 3737 vrele(vp); 3738 return (EEXIST); 3739 } 3740 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3741 NDFREE(&nd, NDF_ONLY_PNBUF); 3742 vput(nd.ni_dvp); 3743 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0) 3744 return (error); 3745 goto restart; 3746 } 3747 VATTR_NULL(&vattr); 3748 vattr.va_type = VDIR; 3749 vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask; 3750 #ifdef MAC 3751 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd, 3752 &vattr); 3753 if (error != 0) 3754 goto out; 3755 #endif 3756 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3757 #ifdef MAC 3758 out: 3759 #endif 3760 NDFREE(&nd, NDF_ONLY_PNBUF); 3761 vput(nd.ni_dvp); 3762 if (error == 0) 3763 vput(nd.ni_vp); 3764 vn_finished_write(mp); 3765 return (error); 3766 } 3767 3768 /* 3769 * Remove a directory file. 3770 */ 3771 #ifndef _SYS_SYSPROTO_H_ 3772 struct rmdir_args { 3773 char *path; 3774 }; 3775 #endif 3776 int 3777 sys_rmdir(struct thread *td, struct rmdir_args *uap) 3778 { 3779 3780 return (kern_frmdirat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE, 3781 0)); 3782 } 3783 3784 int 3785 kern_frmdirat(struct thread *td, int dfd, const char *path, int fd, 3786 enum uio_seg pathseg, int flag) 3787 { 3788 struct mount *mp; 3789 struct vnode *vp; 3790 struct file *fp; 3791 struct nameidata nd; 3792 cap_rights_t rights; 3793 int error; 3794 3795 fp = NULL; 3796 if (fd != FD_NONE) { 3797 error = getvnode(td, fd, cap_rights_init_one(&rights, CAP_LOOKUP), 3798 &fp); 3799 if (error != 0) 3800 return (error); 3801 } 3802 3803 restart: 3804 bwillwrite(); 3805 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 | 3806 ((flag & AT_BENEATH) != 0 ? BENEATH : 0), 3807 pathseg, path, dfd, &cap_unlinkat_rights, td); 3808 if ((error = namei(&nd)) != 0) 3809 goto fdout; 3810 vp = nd.ni_vp; 3811 if (vp->v_type != VDIR) { 3812 error = ENOTDIR; 3813 goto out; 3814 } 3815 /* 3816 * No rmdir "." please. 3817 */ 3818 if (nd.ni_dvp == vp) { 3819 error = EINVAL; 3820 goto out; 3821 } 3822 /* 3823 * The root of a mounted filesystem cannot be deleted. 3824 */ 3825 if (vp->v_vflag & VV_ROOT) { 3826 error = EBUSY; 3827 goto out; 3828 } 3829 3830 if (fp != NULL && fp->f_vnode != vp) { 3831 if (VN_IS_DOOMED(fp->f_vnode)) 3832 error = EBADF; 3833 else 3834 error = EDEADLK; 3835 goto out; 3836 } 3837 3838 #ifdef MAC 3839 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp, 3840 &nd.ni_cnd); 3841 if (error != 0) 3842 goto out; 3843 #endif 3844 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3845 NDFREE(&nd, NDF_ONLY_PNBUF); 3846 vput(vp); 3847 if (nd.ni_dvp == vp) 3848 vrele(nd.ni_dvp); 3849 else 3850 vput(nd.ni_dvp); 3851 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0) 3852 goto fdout; 3853 goto restart; 3854 } 3855 vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK); 3856 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3857 vn_finished_write(mp); 3858 out: 3859 NDFREE(&nd, NDF_ONLY_PNBUF); 3860 vput(vp); 3861 if (nd.ni_dvp == vp) 3862 vrele(nd.ni_dvp); 3863 else 3864 vput(nd.ni_dvp); 3865 fdout: 3866 if (fp != NULL) 3867 fdrop(fp, td); 3868 return (error); 3869 } 3870 3871 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11) 3872 int 3873 freebsd11_kern_getdirentries(struct thread *td, int fd, char *ubuf, u_int count, 3874 long *basep, void (*func)(struct freebsd11_dirent *)) 3875 { 3876 struct freebsd11_dirent dstdp; 3877 struct dirent *dp, *edp; 3878 char *dirbuf; 3879 off_t base; 3880 ssize_t resid, ucount; 3881 int error; 3882 3883 /* XXX arbitrary sanity limit on `count'. */ 3884 count = min(count, 64 * 1024); 3885 3886 dirbuf = malloc(count, M_TEMP, M_WAITOK); 3887 3888 error = kern_getdirentries(td, fd, dirbuf, count, &base, &resid, 3889 UIO_SYSSPACE); 3890 if (error != 0) 3891 goto done; 3892 if (basep != NULL) 3893 *basep = base; 3894 3895 ucount = 0; 3896 for (dp = (struct dirent *)dirbuf, 3897 edp = (struct dirent *)&dirbuf[count - resid]; 3898 ucount < count && dp < edp; ) { 3899 if (dp->d_reclen == 0) 3900 break; 3901 MPASS(dp->d_reclen >= _GENERIC_DIRLEN(0)); 3902 if (dp->d_namlen >= sizeof(dstdp.d_name)) 3903 continue; 3904 dstdp.d_type = dp->d_type; 3905 dstdp.d_namlen = dp->d_namlen; 3906 dstdp.d_fileno = dp->d_fileno; /* truncate */ 3907 if (dstdp.d_fileno != dp->d_fileno) { 3908 switch (ino64_trunc_error) { 3909 default: 3910 case 0: 3911 break; 3912 case 1: 3913 error = EOVERFLOW; 3914 goto done; 3915 case 2: 3916 dstdp.d_fileno = UINT32_MAX; 3917 break; 3918 } 3919 } 3920 dstdp.d_reclen = sizeof(dstdp) - sizeof(dstdp.d_name) + 3921 ((dp->d_namlen + 1 + 3) &~ 3); 3922 bcopy(dp->d_name, dstdp.d_name, dstdp.d_namlen); 3923 bzero(dstdp.d_name + dstdp.d_namlen, 3924 dstdp.d_reclen - offsetof(struct freebsd11_dirent, d_name) - 3925 dstdp.d_namlen); 3926 MPASS(dstdp.d_reclen <= dp->d_reclen); 3927 MPASS(ucount + dstdp.d_reclen <= count); 3928 if (func != NULL) 3929 func(&dstdp); 3930 error = copyout(&dstdp, ubuf + ucount, dstdp.d_reclen); 3931 if (error != 0) 3932 break; 3933 dp = (struct dirent *)((char *)dp + dp->d_reclen); 3934 ucount += dstdp.d_reclen; 3935 } 3936 3937 done: 3938 free(dirbuf, M_TEMP); 3939 if (error == 0) 3940 td->td_retval[0] = ucount; 3941 return (error); 3942 } 3943 #endif /* COMPAT */ 3944 3945 #ifdef COMPAT_43 3946 static void 3947 ogetdirentries_cvt(struct freebsd11_dirent *dp) 3948 { 3949 #if (BYTE_ORDER == LITTLE_ENDIAN) 3950 /* 3951 * The expected low byte of dp->d_namlen is our dp->d_type. 3952 * The high MBZ byte of dp->d_namlen is our dp->d_namlen. 3953 */ 3954 dp->d_type = dp->d_namlen; 3955 dp->d_namlen = 0; 3956 #else 3957 /* 3958 * The dp->d_type is the high byte of the expected dp->d_namlen, 3959 * so must be zero'ed. 3960 */ 3961 dp->d_type = 0; 3962 #endif 3963 } 3964 3965 /* 3966 * Read a block of directory entries in a filesystem independent format. 3967 */ 3968 #ifndef _SYS_SYSPROTO_H_ 3969 struct ogetdirentries_args { 3970 int fd; 3971 char *buf; 3972 u_int count; 3973 long *basep; 3974 }; 3975 #endif 3976 int 3977 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap) 3978 { 3979 long loff; 3980 int error; 3981 3982 error = kern_ogetdirentries(td, uap, &loff); 3983 if (error == 0) 3984 error = copyout(&loff, uap->basep, sizeof(long)); 3985 return (error); 3986 } 3987 3988 int 3989 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap, 3990 long *ploff) 3991 { 3992 long base; 3993 int error; 3994 3995 /* XXX arbitrary sanity limit on `count'. */ 3996 if (uap->count > 64 * 1024) 3997 return (EINVAL); 3998 3999 error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count, 4000 &base, ogetdirentries_cvt); 4001 4002 if (error == 0 && uap->basep != NULL) 4003 error = copyout(&base, uap->basep, sizeof(long)); 4004 4005 return (error); 4006 } 4007 #endif /* COMPAT_43 */ 4008 4009 #if defined(COMPAT_FREEBSD11) 4010 #ifndef _SYS_SYSPROTO_H_ 4011 struct freebsd11_getdirentries_args { 4012 int fd; 4013 char *buf; 4014 u_int count; 4015 long *basep; 4016 }; 4017 #endif 4018 int 4019 freebsd11_getdirentries(struct thread *td, 4020 struct freebsd11_getdirentries_args *uap) 4021 { 4022 long base; 4023 int error; 4024 4025 error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count, 4026 &base, NULL); 4027 4028 if (error == 0 && uap->basep != NULL) 4029 error = copyout(&base, uap->basep, sizeof(long)); 4030 return (error); 4031 } 4032 4033 int 4034 freebsd11_getdents(struct thread *td, struct freebsd11_getdents_args *uap) 4035 { 4036 struct freebsd11_getdirentries_args ap; 4037 4038 ap.fd = uap->fd; 4039 ap.buf = uap->buf; 4040 ap.count = uap->count; 4041 ap.basep = NULL; 4042 return (freebsd11_getdirentries(td, &ap)); 4043 } 4044 #endif /* COMPAT_FREEBSD11 */ 4045 4046 /* 4047 * Read a block of directory entries in a filesystem independent format. 4048 */ 4049 int 4050 sys_getdirentries(struct thread *td, struct getdirentries_args *uap) 4051 { 4052 off_t base; 4053 int error; 4054 4055 error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base, 4056 NULL, UIO_USERSPACE); 4057 if (error != 0) 4058 return (error); 4059 if (uap->basep != NULL) 4060 error = copyout(&base, uap->basep, sizeof(off_t)); 4061 return (error); 4062 } 4063 4064 int 4065 kern_getdirentries(struct thread *td, int fd, char *buf, size_t count, 4066 off_t *basep, ssize_t *residp, enum uio_seg bufseg) 4067 { 4068 struct vnode *vp; 4069 struct file *fp; 4070 struct uio auio; 4071 struct iovec aiov; 4072 off_t loff; 4073 int error, eofflag; 4074 off_t foffset; 4075 4076 AUDIT_ARG_FD(fd); 4077 if (count > IOSIZE_MAX) 4078 return (EINVAL); 4079 auio.uio_resid = count; 4080 error = getvnode(td, fd, &cap_read_rights, &fp); 4081 if (error != 0) 4082 return (error); 4083 if ((fp->f_flag & FREAD) == 0) { 4084 fdrop(fp, td); 4085 return (EBADF); 4086 } 4087 vp = fp->f_vnode; 4088 foffset = foffset_lock(fp, 0); 4089 unionread: 4090 if (vp->v_type != VDIR) { 4091 error = EINVAL; 4092 goto fail; 4093 } 4094 aiov.iov_base = buf; 4095 aiov.iov_len = count; 4096 auio.uio_iov = &aiov; 4097 auio.uio_iovcnt = 1; 4098 auio.uio_rw = UIO_READ; 4099 auio.uio_segflg = bufseg; 4100 auio.uio_td = td; 4101 vn_lock(vp, LK_SHARED | LK_RETRY); 4102 AUDIT_ARG_VNODE1(vp); 4103 loff = auio.uio_offset = foffset; 4104 #ifdef MAC 4105 error = mac_vnode_check_readdir(td->td_ucred, vp); 4106 if (error == 0) 4107 #endif 4108 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL, 4109 NULL); 4110 foffset = auio.uio_offset; 4111 if (error != 0) { 4112 VOP_UNLOCK(vp); 4113 goto fail; 4114 } 4115 if (count == auio.uio_resid && 4116 (vp->v_vflag & VV_ROOT) && 4117 (vp->v_mount->mnt_flag & MNT_UNION)) { 4118 struct vnode *tvp = vp; 4119 4120 vp = vp->v_mount->mnt_vnodecovered; 4121 VREF(vp); 4122 fp->f_vnode = vp; 4123 fp->f_data = vp; 4124 foffset = 0; 4125 vput(tvp); 4126 goto unionread; 4127 } 4128 VOP_UNLOCK(vp); 4129 *basep = loff; 4130 if (residp != NULL) 4131 *residp = auio.uio_resid; 4132 td->td_retval[0] = count - auio.uio_resid; 4133 fail: 4134 foffset_unlock(fp, foffset, 0); 4135 fdrop(fp, td); 4136 return (error); 4137 } 4138 4139 /* 4140 * Set the mode mask for creation of filesystem nodes. 4141 */ 4142 #ifndef _SYS_SYSPROTO_H_ 4143 struct umask_args { 4144 int newmask; 4145 }; 4146 #endif 4147 int 4148 sys_umask(struct thread *td, struct umask_args *uap) 4149 { 4150 struct filedesc *fdp; 4151 4152 fdp = td->td_proc->p_fd; 4153 FILEDESC_XLOCK(fdp); 4154 td->td_retval[0] = fdp->fd_cmask; 4155 fdp->fd_cmask = uap->newmask & ALLPERMS; 4156 FILEDESC_XUNLOCK(fdp); 4157 return (0); 4158 } 4159 4160 /* 4161 * Void all references to file by ripping underlying filesystem away from 4162 * vnode. 4163 */ 4164 #ifndef _SYS_SYSPROTO_H_ 4165 struct revoke_args { 4166 char *path; 4167 }; 4168 #endif 4169 int 4170 sys_revoke(struct thread *td, struct revoke_args *uap) 4171 { 4172 struct vnode *vp; 4173 struct vattr vattr; 4174 struct nameidata nd; 4175 int error; 4176 4177 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE, 4178 uap->path, td); 4179 if ((error = namei(&nd)) != 0) 4180 return (error); 4181 vp = nd.ni_vp; 4182 NDFREE(&nd, NDF_ONLY_PNBUF); 4183 if (vp->v_type != VCHR || vp->v_rdev == NULL) { 4184 error = EINVAL; 4185 goto out; 4186 } 4187 #ifdef MAC 4188 error = mac_vnode_check_revoke(td->td_ucred, vp); 4189 if (error != 0) 4190 goto out; 4191 #endif 4192 error = VOP_GETATTR(vp, &vattr, td->td_ucred); 4193 if (error != 0) 4194 goto out; 4195 if (td->td_ucred->cr_uid != vattr.va_uid) { 4196 error = priv_check(td, PRIV_VFS_ADMIN); 4197 if (error != 0) 4198 goto out; 4199 } 4200 if (vp->v_usecount > 1 || vcount(vp) > 1) 4201 VOP_REVOKE(vp, REVOKEALL); 4202 out: 4203 vput(vp); 4204 return (error); 4205 } 4206 4207 /* 4208 * Convert a user file descriptor to a kernel file entry and check that, if it 4209 * is a capability, the correct rights are present. A reference on the file 4210 * entry is held upon returning. 4211 */ 4212 int 4213 getvnode(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp) 4214 { 4215 struct file *fp; 4216 int error; 4217 4218 error = fget_unlocked(td->td_proc->p_fd, fd, rightsp, &fp); 4219 if (error != 0) 4220 return (error); 4221 4222 /* 4223 * The file could be not of the vnode type, or it may be not 4224 * yet fully initialized, in which case the f_vnode pointer 4225 * may be set, but f_ops is still badfileops. E.g., 4226 * devfs_open() transiently create such situation to 4227 * facilitate csw d_fdopen(). 4228 * 4229 * Dupfdopen() handling in kern_openat() installs the 4230 * half-baked file into the process descriptor table, allowing 4231 * other thread to dereference it. Guard against the race by 4232 * checking f_ops. 4233 */ 4234 if (fp->f_vnode == NULL || fp->f_ops == &badfileops) { 4235 fdrop(fp, td); 4236 return (EINVAL); 4237 } 4238 *fpp = fp; 4239 return (0); 4240 } 4241 4242 /* 4243 * Get an (NFS) file handle. 4244 */ 4245 #ifndef _SYS_SYSPROTO_H_ 4246 struct lgetfh_args { 4247 char *fname; 4248 fhandle_t *fhp; 4249 }; 4250 #endif 4251 int 4252 sys_lgetfh(struct thread *td, struct lgetfh_args *uap) 4253 { 4254 4255 return (kern_getfhat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->fname, 4256 UIO_USERSPACE, uap->fhp)); 4257 } 4258 4259 #ifndef _SYS_SYSPROTO_H_ 4260 struct getfh_args { 4261 char *fname; 4262 fhandle_t *fhp; 4263 }; 4264 #endif 4265 int 4266 sys_getfh(struct thread *td, struct getfh_args *uap) 4267 { 4268 4269 return (kern_getfhat(td, 0, AT_FDCWD, uap->fname, UIO_USERSPACE, 4270 uap->fhp)); 4271 } 4272 4273 /* 4274 * syscall for the rpc.lockd to use to translate an open descriptor into 4275 * a NFS file handle. 4276 * 4277 * warning: do not remove the priv_check() call or this becomes one giant 4278 * security hole. 4279 */ 4280 #ifndef _SYS_SYSPROTO_H_ 4281 struct getfhat_args { 4282 int fd; 4283 char *path; 4284 fhandle_t *fhp; 4285 int flags; 4286 }; 4287 #endif 4288 int 4289 sys_getfhat(struct thread *td, struct getfhat_args *uap) 4290 { 4291 4292 if ((uap->flags & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH)) != 0) 4293 return (EINVAL); 4294 return (kern_getfhat(td, uap->flags, uap->fd, uap->path, UIO_USERSPACE, 4295 uap->fhp)); 4296 } 4297 4298 static int 4299 kern_getfhat(struct thread *td, int flags, int fd, const char *path, 4300 enum uio_seg pathseg, fhandle_t *fhp) 4301 { 4302 struct nameidata nd; 4303 fhandle_t fh; 4304 struct vnode *vp; 4305 int error; 4306 4307 error = priv_check(td, PRIV_VFS_GETFH); 4308 if (error != 0) 4309 return (error); 4310 NDINIT_AT(&nd, LOOKUP, ((flags & AT_SYMLINK_NOFOLLOW) != 0 ? NOFOLLOW : 4311 FOLLOW) | ((flags & AT_BENEATH) != 0 ? BENEATH : 0) | LOCKLEAF | 4312 AUDITVNODE1, pathseg, path, fd, td); 4313 error = namei(&nd); 4314 if (error != 0) 4315 return (error); 4316 NDFREE(&nd, NDF_ONLY_PNBUF); 4317 vp = nd.ni_vp; 4318 bzero(&fh, sizeof(fh)); 4319 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid; 4320 error = VOP_VPTOFH(vp, &fh.fh_fid); 4321 vput(vp); 4322 if (error == 0) 4323 error = copyout(&fh, fhp, sizeof (fh)); 4324 return (error); 4325 } 4326 4327 #ifndef _SYS_SYSPROTO_H_ 4328 struct fhlink_args { 4329 fhandle_t *fhp; 4330 const char *to; 4331 }; 4332 #endif 4333 int 4334 sys_fhlink(struct thread *td, struct fhlink_args *uap) 4335 { 4336 4337 return (kern_fhlinkat(td, AT_FDCWD, uap->to, UIO_USERSPACE, uap->fhp)); 4338 } 4339 4340 #ifndef _SYS_SYSPROTO_H_ 4341 struct fhlinkat_args { 4342 fhandle_t *fhp; 4343 int tofd; 4344 const char *to; 4345 }; 4346 #endif 4347 int 4348 sys_fhlinkat(struct thread *td, struct fhlinkat_args *uap) 4349 { 4350 4351 return (kern_fhlinkat(td, uap->tofd, uap->to, UIO_USERSPACE, uap->fhp)); 4352 } 4353 4354 static int 4355 kern_fhlinkat(struct thread *td, int fd, const char *path, 4356 enum uio_seg pathseg, fhandle_t *fhp) 4357 { 4358 fhandle_t fh; 4359 struct mount *mp; 4360 struct vnode *vp; 4361 int error; 4362 4363 error = priv_check(td, PRIV_VFS_GETFH); 4364 if (error != 0) 4365 return (error); 4366 error = copyin(fhp, &fh, sizeof(fh)); 4367 if (error != 0) 4368 return (error); 4369 do { 4370 bwillwrite(); 4371 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL) 4372 return (ESTALE); 4373 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp); 4374 vfs_unbusy(mp); 4375 if (error != 0) 4376 return (error); 4377 VOP_UNLOCK(vp); 4378 } while ((error = kern_linkat_vp(td, vp, fd, path, pathseg)) == EAGAIN); 4379 return (error); 4380 } 4381 4382 #ifndef _SYS_SYSPROTO_H_ 4383 struct fhreadlink_args { 4384 fhandle_t *fhp; 4385 char *buf; 4386 size_t bufsize; 4387 }; 4388 #endif 4389 int 4390 sys_fhreadlink(struct thread *td, struct fhreadlink_args *uap) 4391 { 4392 fhandle_t fh; 4393 struct mount *mp; 4394 struct vnode *vp; 4395 int error; 4396 4397 error = priv_check(td, PRIV_VFS_GETFH); 4398 if (error != 0) 4399 return (error); 4400 if (uap->bufsize > IOSIZE_MAX) 4401 return (EINVAL); 4402 error = copyin(uap->fhp, &fh, sizeof(fh)); 4403 if (error != 0) 4404 return (error); 4405 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL) 4406 return (ESTALE); 4407 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp); 4408 vfs_unbusy(mp); 4409 if (error != 0) 4410 return (error); 4411 error = kern_readlink_vp(vp, uap->buf, UIO_USERSPACE, uap->bufsize, td); 4412 vput(vp); 4413 return (error); 4414 } 4415 4416 /* 4417 * syscall for the rpc.lockd to use to translate a NFS file handle into an 4418 * open descriptor. 4419 * 4420 * warning: do not remove the priv_check() call or this becomes one giant 4421 * security hole. 4422 */ 4423 #ifndef _SYS_SYSPROTO_H_ 4424 struct fhopen_args { 4425 const struct fhandle *u_fhp; 4426 int flags; 4427 }; 4428 #endif 4429 int 4430 sys_fhopen(struct thread *td, struct fhopen_args *uap) 4431 { 4432 struct mount *mp; 4433 struct vnode *vp; 4434 struct fhandle fhp; 4435 struct file *fp; 4436 int fmode, error; 4437 int indx; 4438 4439 error = priv_check(td, PRIV_VFS_FHOPEN); 4440 if (error != 0) 4441 return (error); 4442 indx = -1; 4443 fmode = FFLAGS(uap->flags); 4444 /* why not allow a non-read/write open for our lockd? */ 4445 if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT)) 4446 return (EINVAL); 4447 error = copyin(uap->u_fhp, &fhp, sizeof(fhp)); 4448 if (error != 0) 4449 return(error); 4450 /* find the mount point */ 4451 mp = vfs_busyfs(&fhp.fh_fsid); 4452 if (mp == NULL) 4453 return (ESTALE); 4454 /* now give me my vnode, it gets returned to me locked */ 4455 error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp); 4456 vfs_unbusy(mp); 4457 if (error != 0) 4458 return (error); 4459 4460 error = falloc_noinstall(td, &fp); 4461 if (error != 0) { 4462 vput(vp); 4463 return (error); 4464 } 4465 /* 4466 * An extra reference on `fp' has been held for us by 4467 * falloc_noinstall(). 4468 */ 4469 4470 #ifdef INVARIANTS 4471 td->td_dupfd = -1; 4472 #endif 4473 error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp); 4474 if (error != 0) { 4475 KASSERT(fp->f_ops == &badfileops, 4476 ("VOP_OPEN in fhopen() set f_ops")); 4477 KASSERT(td->td_dupfd < 0, 4478 ("fhopen() encountered fdopen()")); 4479 4480 vput(vp); 4481 goto bad; 4482 } 4483 #ifdef INVARIANTS 4484 td->td_dupfd = 0; 4485 #endif 4486 fp->f_vnode = vp; 4487 fp->f_seqcount[UIO_READ] = 1; 4488 fp->f_seqcount[UIO_WRITE] = 1; 4489 finit(fp, (fmode & FMASK) | (fp->f_flag & FHASLOCK), DTYPE_VNODE, vp, 4490 &vnops); 4491 VOP_UNLOCK(vp); 4492 if ((fmode & O_TRUNC) != 0) { 4493 error = fo_truncate(fp, 0, td->td_ucred, td); 4494 if (error != 0) 4495 goto bad; 4496 } 4497 4498 error = finstall(td, fp, &indx, fmode, NULL); 4499 bad: 4500 fdrop(fp, td); 4501 td->td_retval[0] = indx; 4502 return (error); 4503 } 4504 4505 /* 4506 * Stat an (NFS) file handle. 4507 */ 4508 #ifndef _SYS_SYSPROTO_H_ 4509 struct fhstat_args { 4510 struct fhandle *u_fhp; 4511 struct stat *sb; 4512 }; 4513 #endif 4514 int 4515 sys_fhstat(struct thread *td, struct fhstat_args *uap) 4516 { 4517 struct stat sb; 4518 struct fhandle fh; 4519 int error; 4520 4521 error = copyin(uap->u_fhp, &fh, sizeof(fh)); 4522 if (error != 0) 4523 return (error); 4524 error = kern_fhstat(td, fh, &sb); 4525 if (error == 0) 4526 error = copyout(&sb, uap->sb, sizeof(sb)); 4527 return (error); 4528 } 4529 4530 int 4531 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb) 4532 { 4533 struct mount *mp; 4534 struct vnode *vp; 4535 int error; 4536 4537 error = priv_check(td, PRIV_VFS_FHSTAT); 4538 if (error != 0) 4539 return (error); 4540 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL) 4541 return (ESTALE); 4542 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp); 4543 vfs_unbusy(mp); 4544 if (error != 0) 4545 return (error); 4546 error = vn_stat(vp, sb, td->td_ucred, NOCRED, td); 4547 vput(vp); 4548 return (error); 4549 } 4550 4551 /* 4552 * Implement fstatfs() for (NFS) file handles. 4553 */ 4554 #ifndef _SYS_SYSPROTO_H_ 4555 struct fhstatfs_args { 4556 struct fhandle *u_fhp; 4557 struct statfs *buf; 4558 }; 4559 #endif 4560 int 4561 sys_fhstatfs(struct thread *td, struct fhstatfs_args *uap) 4562 { 4563 struct statfs *sfp; 4564 fhandle_t fh; 4565 int error; 4566 4567 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t)); 4568 if (error != 0) 4569 return (error); 4570 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 4571 error = kern_fhstatfs(td, fh, sfp); 4572 if (error == 0) 4573 error = copyout(sfp, uap->buf, sizeof(*sfp)); 4574 free(sfp, M_STATFS); 4575 return (error); 4576 } 4577 4578 int 4579 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf) 4580 { 4581 struct mount *mp; 4582 struct vnode *vp; 4583 int error; 4584 4585 error = priv_check(td, PRIV_VFS_FHSTATFS); 4586 if (error != 0) 4587 return (error); 4588 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL) 4589 return (ESTALE); 4590 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp); 4591 if (error != 0) { 4592 vfs_unbusy(mp); 4593 return (error); 4594 } 4595 vput(vp); 4596 error = prison_canseemount(td->td_ucred, mp); 4597 if (error != 0) 4598 goto out; 4599 #ifdef MAC 4600 error = mac_mount_check_stat(td->td_ucred, mp); 4601 if (error != 0) 4602 goto out; 4603 #endif 4604 error = VFS_STATFS(mp, buf); 4605 out: 4606 vfs_unbusy(mp); 4607 return (error); 4608 } 4609 4610 /* 4611 * Unlike madvise(2), we do not make a best effort to remember every 4612 * possible caching hint. Instead, we remember the last setting with 4613 * the exception that we will allow POSIX_FADV_NORMAL to adjust the 4614 * region of any current setting. 4615 */ 4616 int 4617 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len, 4618 int advice) 4619 { 4620 struct fadvise_info *fa, *new; 4621 struct file *fp; 4622 struct vnode *vp; 4623 off_t end; 4624 int error; 4625 4626 if (offset < 0 || len < 0 || offset > OFF_MAX - len) 4627 return (EINVAL); 4628 AUDIT_ARG_VALUE(advice); 4629 switch (advice) { 4630 case POSIX_FADV_SEQUENTIAL: 4631 case POSIX_FADV_RANDOM: 4632 case POSIX_FADV_NOREUSE: 4633 new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK); 4634 break; 4635 case POSIX_FADV_NORMAL: 4636 case POSIX_FADV_WILLNEED: 4637 case POSIX_FADV_DONTNEED: 4638 new = NULL; 4639 break; 4640 default: 4641 return (EINVAL); 4642 } 4643 /* XXX: CAP_POSIX_FADVISE? */ 4644 AUDIT_ARG_FD(fd); 4645 error = fget(td, fd, &cap_no_rights, &fp); 4646 if (error != 0) 4647 goto out; 4648 AUDIT_ARG_FILE(td->td_proc, fp); 4649 if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) { 4650 error = ESPIPE; 4651 goto out; 4652 } 4653 if (fp->f_type != DTYPE_VNODE) { 4654 error = ENODEV; 4655 goto out; 4656 } 4657 vp = fp->f_vnode; 4658 if (vp->v_type != VREG) { 4659 error = ENODEV; 4660 goto out; 4661 } 4662 if (len == 0) 4663 end = OFF_MAX; 4664 else 4665 end = offset + len - 1; 4666 switch (advice) { 4667 case POSIX_FADV_SEQUENTIAL: 4668 case POSIX_FADV_RANDOM: 4669 case POSIX_FADV_NOREUSE: 4670 /* 4671 * Try to merge any existing non-standard region with 4672 * this new region if possible, otherwise create a new 4673 * non-standard region for this request. 4674 */ 4675 mtx_pool_lock(mtxpool_sleep, fp); 4676 fa = fp->f_advice; 4677 if (fa != NULL && fa->fa_advice == advice && 4678 ((fa->fa_start <= end && fa->fa_end >= offset) || 4679 (end != OFF_MAX && fa->fa_start == end + 1) || 4680 (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) { 4681 if (offset < fa->fa_start) 4682 fa->fa_start = offset; 4683 if (end > fa->fa_end) 4684 fa->fa_end = end; 4685 } else { 4686 new->fa_advice = advice; 4687 new->fa_start = offset; 4688 new->fa_end = end; 4689 fp->f_advice = new; 4690 new = fa; 4691 } 4692 mtx_pool_unlock(mtxpool_sleep, fp); 4693 break; 4694 case POSIX_FADV_NORMAL: 4695 /* 4696 * If a the "normal" region overlaps with an existing 4697 * non-standard region, trim or remove the 4698 * non-standard region. 4699 */ 4700 mtx_pool_lock(mtxpool_sleep, fp); 4701 fa = fp->f_advice; 4702 if (fa != NULL) { 4703 if (offset <= fa->fa_start && end >= fa->fa_end) { 4704 new = fa; 4705 fp->f_advice = NULL; 4706 } else if (offset <= fa->fa_start && 4707 end >= fa->fa_start) 4708 fa->fa_start = end + 1; 4709 else if (offset <= fa->fa_end && end >= fa->fa_end) 4710 fa->fa_end = offset - 1; 4711 else if (offset >= fa->fa_start && end <= fa->fa_end) { 4712 /* 4713 * If the "normal" region is a middle 4714 * portion of the existing 4715 * non-standard region, just remove 4716 * the whole thing rather than picking 4717 * one side or the other to 4718 * preserve. 4719 */ 4720 new = fa; 4721 fp->f_advice = NULL; 4722 } 4723 } 4724 mtx_pool_unlock(mtxpool_sleep, fp); 4725 break; 4726 case POSIX_FADV_WILLNEED: 4727 case POSIX_FADV_DONTNEED: 4728 error = VOP_ADVISE(vp, offset, end, advice); 4729 break; 4730 } 4731 out: 4732 if (fp != NULL) 4733 fdrop(fp, td); 4734 free(new, M_FADVISE); 4735 return (error); 4736 } 4737 4738 int 4739 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap) 4740 { 4741 int error; 4742 4743 error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len, 4744 uap->advice); 4745 return (kern_posix_error(td, error)); 4746 } 4747 4748 int 4749 kern_copy_file_range(struct thread *td, int infd, off_t *inoffp, int outfd, 4750 off_t *outoffp, size_t len, unsigned int flags) 4751 { 4752 struct file *infp, *outfp; 4753 struct vnode *invp, *outvp; 4754 int error; 4755 size_t retlen; 4756 void *rl_rcookie, *rl_wcookie; 4757 off_t savinoff, savoutoff; 4758 4759 infp = outfp = NULL; 4760 rl_rcookie = rl_wcookie = NULL; 4761 savinoff = -1; 4762 error = 0; 4763 retlen = 0; 4764 4765 if (flags != 0) { 4766 error = EINVAL; 4767 goto out; 4768 } 4769 if (len > SSIZE_MAX) 4770 /* 4771 * Although the len argument is size_t, the return argument 4772 * is ssize_t (which is signed). Therefore a size that won't 4773 * fit in ssize_t can't be returned. 4774 */ 4775 len = SSIZE_MAX; 4776 4777 /* Get the file structures for the file descriptors. */ 4778 error = fget_read(td, infd, &cap_read_rights, &infp); 4779 if (error != 0) 4780 goto out; 4781 if (infp->f_ops == &badfileops) { 4782 error = EBADF; 4783 goto out; 4784 } 4785 if (infp->f_vnode == NULL) { 4786 error = EINVAL; 4787 goto out; 4788 } 4789 error = fget_write(td, outfd, &cap_write_rights, &outfp); 4790 if (error != 0) 4791 goto out; 4792 if (outfp->f_ops == &badfileops) { 4793 error = EBADF; 4794 goto out; 4795 } 4796 if (outfp->f_vnode == NULL) { 4797 error = EINVAL; 4798 goto out; 4799 } 4800 4801 /* Set the offset pointers to the correct place. */ 4802 if (inoffp == NULL) 4803 inoffp = &infp->f_offset; 4804 if (outoffp == NULL) 4805 outoffp = &outfp->f_offset; 4806 savinoff = *inoffp; 4807 savoutoff = *outoffp; 4808 4809 invp = infp->f_vnode; 4810 outvp = outfp->f_vnode; 4811 /* Sanity check the f_flag bits. */ 4812 if ((outfp->f_flag & (FWRITE | FAPPEND)) != FWRITE || 4813 (infp->f_flag & FREAD) == 0) { 4814 error = EBADF; 4815 goto out; 4816 } 4817 4818 /* If len == 0, just return 0. */ 4819 if (len == 0) 4820 goto out; 4821 4822 /* 4823 * If infp and outfp refer to the same file, the byte ranges cannot 4824 * overlap. 4825 */ 4826 if (invp == outvp && ((savinoff <= savoutoff && savinoff + len > 4827 savoutoff) || (savinoff > savoutoff && savoutoff + len > 4828 savinoff))) { 4829 error = EINVAL; 4830 goto out; 4831 } 4832 4833 /* Range lock the byte ranges for both invp and outvp. */ 4834 for (;;) { 4835 rl_wcookie = vn_rangelock_wlock(outvp, *outoffp, *outoffp + 4836 len); 4837 rl_rcookie = vn_rangelock_tryrlock(invp, *inoffp, *inoffp + 4838 len); 4839 if (rl_rcookie != NULL) 4840 break; 4841 vn_rangelock_unlock(outvp, rl_wcookie); 4842 rl_rcookie = vn_rangelock_rlock(invp, *inoffp, *inoffp + len); 4843 vn_rangelock_unlock(invp, rl_rcookie); 4844 } 4845 4846 retlen = len; 4847 error = vn_copy_file_range(invp, inoffp, outvp, outoffp, &retlen, 4848 flags, infp->f_cred, outfp->f_cred, td); 4849 out: 4850 if (rl_rcookie != NULL) 4851 vn_rangelock_unlock(invp, rl_rcookie); 4852 if (rl_wcookie != NULL) 4853 vn_rangelock_unlock(outvp, rl_wcookie); 4854 if (savinoff != -1 && (error == EINTR || error == ERESTART)) { 4855 *inoffp = savinoff; 4856 *outoffp = savoutoff; 4857 } 4858 if (outfp != NULL) 4859 fdrop(outfp, td); 4860 if (infp != NULL) 4861 fdrop(infp, td); 4862 td->td_retval[0] = retlen; 4863 return (error); 4864 } 4865 4866 int 4867 sys_copy_file_range(struct thread *td, struct copy_file_range_args *uap) 4868 { 4869 off_t inoff, outoff, *inoffp, *outoffp; 4870 int error; 4871 4872 inoffp = outoffp = NULL; 4873 if (uap->inoffp != NULL) { 4874 error = copyin(uap->inoffp, &inoff, sizeof(off_t)); 4875 if (error != 0) 4876 return (error); 4877 inoffp = &inoff; 4878 } 4879 if (uap->outoffp != NULL) { 4880 error = copyin(uap->outoffp, &outoff, sizeof(off_t)); 4881 if (error != 0) 4882 return (error); 4883 outoffp = &outoff; 4884 } 4885 error = kern_copy_file_range(td, uap->infd, inoffp, uap->outfd, 4886 outoffp, uap->len, uap->flags); 4887 if (error == 0 && uap->inoffp != NULL) 4888 error = copyout(inoffp, uap->inoffp, sizeof(off_t)); 4889 if (error == 0 && uap->outoffp != NULL) 4890 error = copyout(outoffp, uap->outoffp, sizeof(off_t)); 4891 return (error); 4892 } 4893