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_compat.h" 44 #include "opt_ktrace.h" 45 46 #include <sys/param.h> 47 #include <sys/systm.h> 48 #include <sys/bio.h> 49 #include <sys/buf.h> 50 #include <sys/capsicum.h> 51 #include <sys/disk.h> 52 #include <sys/sysent.h> 53 #include <sys/malloc.h> 54 #include <sys/mount.h> 55 #include <sys/mutex.h> 56 #include <sys/sysproto.h> 57 #include <sys/namei.h> 58 #include <sys/filedesc.h> 59 #include <sys/kernel.h> 60 #include <sys/fcntl.h> 61 #include <sys/file.h> 62 #include <sys/filio.h> 63 #include <sys/limits.h> 64 #include <sys/linker.h> 65 #include <sys/rwlock.h> 66 #include <sys/sdt.h> 67 #include <sys/stat.h> 68 #include <sys/sx.h> 69 #include <sys/unistd.h> 70 #include <sys/vnode.h> 71 #include <sys/priv.h> 72 #include <sys/proc.h> 73 #include <sys/dirent.h> 74 #include <sys/jail.h> 75 #include <sys/syscallsubr.h> 76 #include <sys/sysctl.h> 77 #ifdef KTRACE 78 #include <sys/ktrace.h> 79 #endif 80 81 #include <machine/stdarg.h> 82 83 #include <security/audit/audit.h> 84 #include <security/mac/mac_framework.h> 85 86 #include <vm/vm.h> 87 #include <vm/vm_object.h> 88 #include <vm/vm_page.h> 89 #include <vm/uma.h> 90 91 #include <ufs/ufs/quota.h> 92 93 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information"); 94 95 SDT_PROVIDER_DEFINE(vfs); 96 SDT_PROBE_DEFINE2(vfs, , stat, mode, "char *", "int"); 97 SDT_PROBE_DEFINE2(vfs, , stat, reg, "char *", "int"); 98 99 static int kern_chflagsat(struct thread *td, int fd, const char *path, 100 enum uio_seg pathseg, u_long flags, int atflag); 101 static int setfflags(struct thread *td, struct vnode *, u_long); 102 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *); 103 static int getutimens(const struct timespec *, enum uio_seg, 104 struct timespec *, int *); 105 static int setutimes(struct thread *td, struct vnode *, 106 const struct timespec *, int, int); 107 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred, 108 struct thread *td); 109 110 /* 111 * Sync each mounted filesystem. 112 */ 113 #ifndef _SYS_SYSPROTO_H_ 114 struct sync_args { 115 int dummy; 116 }; 117 #endif 118 /* ARGSUSED */ 119 int 120 sys_sync(struct thread *td, struct sync_args *uap) 121 { 122 struct mount *mp, *nmp; 123 int save; 124 125 mtx_lock(&mountlist_mtx); 126 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) { 127 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) { 128 nmp = TAILQ_NEXT(mp, mnt_list); 129 continue; 130 } 131 if ((mp->mnt_flag & MNT_RDONLY) == 0 && 132 vn_start_write(NULL, &mp, V_NOWAIT) == 0) { 133 save = curthread_pflags_set(TDP_SYNCIO); 134 vfs_msync(mp, MNT_NOWAIT); 135 VFS_SYNC(mp, MNT_NOWAIT); 136 curthread_pflags_restore(save); 137 vn_finished_write(mp); 138 } 139 mtx_lock(&mountlist_mtx); 140 nmp = TAILQ_NEXT(mp, mnt_list); 141 vfs_unbusy(mp); 142 } 143 mtx_unlock(&mountlist_mtx); 144 return (0); 145 } 146 147 /* 148 * Change filesystem quotas. 149 */ 150 #ifndef _SYS_SYSPROTO_H_ 151 struct quotactl_args { 152 char *path; 153 int cmd; 154 int uid; 155 caddr_t arg; 156 }; 157 #endif 158 int 159 sys_quotactl(struct thread *td, struct quotactl_args *uap) 160 { 161 struct mount *mp; 162 struct nameidata nd; 163 int error; 164 165 AUDIT_ARG_CMD(uap->cmd); 166 AUDIT_ARG_UID(uap->uid); 167 if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS)) 168 return (EPERM); 169 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE, 170 uap->path, td); 171 if ((error = namei(&nd)) != 0) 172 return (error); 173 NDFREE(&nd, NDF_ONLY_PNBUF); 174 mp = nd.ni_vp->v_mount; 175 vfs_ref(mp); 176 vput(nd.ni_vp); 177 error = vfs_busy(mp, 0); 178 vfs_rel(mp); 179 if (error != 0) 180 return (error); 181 error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg); 182 183 /* 184 * Since quota on operation typically needs to open quota 185 * file, the Q_QUOTAON handler needs to unbusy the mount point 186 * before calling into namei. Otherwise, unmount might be 187 * started between two vfs_busy() invocations (first is our, 188 * second is from mount point cross-walk code in lookup()), 189 * causing deadlock. 190 * 191 * Require that Q_QUOTAON handles the vfs_busy() reference on 192 * its own, always returning with ubusied mount point. 193 */ 194 if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON) 195 vfs_unbusy(mp); 196 return (error); 197 } 198 199 /* 200 * Used by statfs conversion routines to scale the block size up if 201 * necessary so that all of the block counts are <= 'max_size'. Note 202 * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero 203 * value of 'n'. 204 */ 205 void 206 statfs_scale_blocks(struct statfs *sf, long max_size) 207 { 208 uint64_t count; 209 int shift; 210 211 KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__)); 212 213 /* 214 * Attempt to scale the block counts to give a more accurate 215 * overview to userland of the ratio of free space to used 216 * space. To do this, find the largest block count and compute 217 * a divisor that lets it fit into a signed integer <= max_size. 218 */ 219 if (sf->f_bavail < 0) 220 count = -sf->f_bavail; 221 else 222 count = sf->f_bavail; 223 count = MAX(sf->f_blocks, MAX(sf->f_bfree, count)); 224 if (count <= max_size) 225 return; 226 227 count >>= flsl(max_size); 228 shift = 0; 229 while (count > 0) { 230 shift++; 231 count >>=1; 232 } 233 234 sf->f_bsize <<= shift; 235 sf->f_blocks >>= shift; 236 sf->f_bfree >>= shift; 237 sf->f_bavail >>= shift; 238 } 239 240 static int 241 kern_do_statfs(struct thread *td, struct mount *mp, struct statfs *buf) 242 { 243 struct statfs *sp; 244 int error; 245 246 if (mp == NULL) 247 return (EBADF); 248 error = vfs_busy(mp, 0); 249 vfs_rel(mp); 250 if (error != 0) 251 return (error); 252 #ifdef MAC 253 error = mac_mount_check_stat(td->td_ucred, mp); 254 if (error != 0) 255 goto out; 256 #endif 257 /* 258 * Set these in case the underlying filesystem fails to do so. 259 */ 260 sp = &mp->mnt_stat; 261 sp->f_version = STATFS_VERSION; 262 sp->f_namemax = NAME_MAX; 263 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK; 264 error = VFS_STATFS(mp, sp); 265 if (error != 0) 266 goto out; 267 *buf = *sp; 268 if (priv_check(td, PRIV_VFS_GENERATION)) { 269 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0; 270 prison_enforce_statfs(td->td_ucred, mp, buf); 271 } 272 out: 273 vfs_unbusy(mp); 274 return (error); 275 } 276 277 /* 278 * Get filesystem statistics. 279 */ 280 #ifndef _SYS_SYSPROTO_H_ 281 struct statfs_args { 282 char *path; 283 struct statfs *buf; 284 }; 285 #endif 286 int 287 sys_statfs(struct thread *td, struct statfs_args *uap) 288 { 289 struct statfs *sfp; 290 int error; 291 292 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 293 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp); 294 if (error == 0) 295 error = copyout(sfp, uap->buf, sizeof(struct statfs)); 296 free(sfp, M_STATFS); 297 return (error); 298 } 299 300 int 301 kern_statfs(struct thread *td, char *path, enum uio_seg pathseg, 302 struct statfs *buf) 303 { 304 struct mount *mp; 305 struct nameidata nd; 306 int error; 307 308 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1, 309 pathseg, path, td); 310 error = namei(&nd); 311 if (error != 0) 312 return (error); 313 mp = nd.ni_vp->v_mount; 314 vfs_ref(mp); 315 NDFREE(&nd, NDF_ONLY_PNBUF); 316 vput(nd.ni_vp); 317 return (kern_do_statfs(td, mp, buf)); 318 } 319 320 /* 321 * Get filesystem statistics. 322 */ 323 #ifndef _SYS_SYSPROTO_H_ 324 struct fstatfs_args { 325 int fd; 326 struct statfs *buf; 327 }; 328 #endif 329 int 330 sys_fstatfs(struct thread *td, struct fstatfs_args *uap) 331 { 332 struct statfs *sfp; 333 int error; 334 335 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 336 error = kern_fstatfs(td, uap->fd, sfp); 337 if (error == 0) 338 error = copyout(sfp, uap->buf, sizeof(struct statfs)); 339 free(sfp, M_STATFS); 340 return (error); 341 } 342 343 int 344 kern_fstatfs(struct thread *td, int fd, struct statfs *buf) 345 { 346 struct file *fp; 347 struct mount *mp; 348 struct vnode *vp; 349 cap_rights_t rights; 350 int error; 351 352 AUDIT_ARG_FD(fd); 353 error = getvnode(td, fd, cap_rights_init(&rights, CAP_FSTATFS), &fp); 354 if (error != 0) 355 return (error); 356 vp = fp->f_vnode; 357 vn_lock(vp, LK_SHARED | LK_RETRY); 358 #ifdef AUDIT 359 AUDIT_ARG_VNODE1(vp); 360 #endif 361 mp = vp->v_mount; 362 if (mp != NULL) 363 vfs_ref(mp); 364 VOP_UNLOCK(vp, 0); 365 fdrop(fp, td); 366 return (kern_do_statfs(td, mp, buf)); 367 } 368 369 /* 370 * Get statistics on all filesystems. 371 */ 372 #ifndef _SYS_SYSPROTO_H_ 373 struct getfsstat_args { 374 struct statfs *buf; 375 long bufsize; 376 int mode; 377 }; 378 #endif 379 int 380 sys_getfsstat(struct thread *td, struct getfsstat_args *uap) 381 { 382 size_t count; 383 int error; 384 385 if (uap->bufsize < 0 || uap->bufsize > SIZE_MAX) 386 return (EINVAL); 387 error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count, 388 UIO_USERSPACE, uap->mode); 389 if (error == 0) 390 td->td_retval[0] = count; 391 return (error); 392 } 393 394 /* 395 * If (bufsize > 0 && bufseg == UIO_SYSSPACE) 396 * The caller is responsible for freeing memory which will be allocated 397 * in '*buf'. 398 */ 399 int 400 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize, 401 size_t *countp, enum uio_seg bufseg, int mode) 402 { 403 struct mount *mp, *nmp; 404 struct statfs *sfsp, *sp, *sptmp, *tofree; 405 size_t count, maxcount; 406 int error; 407 408 switch (mode) { 409 case MNT_WAIT: 410 case MNT_NOWAIT: 411 break; 412 default: 413 if (bufseg == UIO_SYSSPACE) 414 *buf = NULL; 415 return (EINVAL); 416 } 417 restart: 418 maxcount = bufsize / sizeof(struct statfs); 419 if (bufsize == 0) { 420 sfsp = NULL; 421 tofree = NULL; 422 } else if (bufseg == UIO_USERSPACE) { 423 sfsp = *buf; 424 tofree = NULL; 425 } else /* if (bufseg == UIO_SYSSPACE) */ { 426 count = 0; 427 mtx_lock(&mountlist_mtx); 428 TAILQ_FOREACH(mp, &mountlist, mnt_list) { 429 count++; 430 } 431 mtx_unlock(&mountlist_mtx); 432 if (maxcount > count) 433 maxcount = count; 434 tofree = sfsp = *buf = malloc(maxcount * sizeof(struct statfs), 435 M_STATFS, M_WAITOK); 436 } 437 count = 0; 438 mtx_lock(&mountlist_mtx); 439 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) { 440 if (prison_canseemount(td->td_ucred, mp) != 0) { 441 nmp = TAILQ_NEXT(mp, mnt_list); 442 continue; 443 } 444 #ifdef MAC 445 if (mac_mount_check_stat(td->td_ucred, mp) != 0) { 446 nmp = TAILQ_NEXT(mp, mnt_list); 447 continue; 448 } 449 #endif 450 if (mode == MNT_WAIT) { 451 if (vfs_busy(mp, MBF_MNTLSTLOCK) != 0) { 452 /* 453 * If vfs_busy() failed, and MBF_NOWAIT 454 * wasn't passed, then the mp is gone. 455 * Furthermore, because of MBF_MNTLSTLOCK, 456 * the mountlist_mtx was dropped. We have 457 * no other choice than to start over. 458 */ 459 mtx_unlock(&mountlist_mtx); 460 free(tofree, M_STATFS); 461 goto restart; 462 } 463 } else { 464 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK) != 0) { 465 nmp = TAILQ_NEXT(mp, mnt_list); 466 continue; 467 } 468 } 469 if (sfsp != NULL && count < maxcount) { 470 sp = &mp->mnt_stat; 471 /* 472 * Set these in case the underlying filesystem 473 * fails to do so. 474 */ 475 sp->f_version = STATFS_VERSION; 476 sp->f_namemax = NAME_MAX; 477 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK; 478 /* 479 * If MNT_NOWAIT is specified, do not refresh 480 * the fsstat cache. 481 */ 482 if (mode != MNT_NOWAIT) { 483 error = VFS_STATFS(mp, sp); 484 if (error != 0) { 485 mtx_lock(&mountlist_mtx); 486 nmp = TAILQ_NEXT(mp, mnt_list); 487 vfs_unbusy(mp); 488 continue; 489 } 490 } 491 if (priv_check(td, PRIV_VFS_GENERATION)) { 492 sptmp = malloc(sizeof(struct statfs), M_STATFS, 493 M_WAITOK); 494 *sptmp = *sp; 495 sptmp->f_fsid.val[0] = sptmp->f_fsid.val[1] = 0; 496 prison_enforce_statfs(td->td_ucred, mp, sptmp); 497 sp = sptmp; 498 } else 499 sptmp = NULL; 500 if (bufseg == UIO_SYSSPACE) { 501 bcopy(sp, sfsp, sizeof(*sp)); 502 free(sptmp, M_STATFS); 503 } else /* if (bufseg == UIO_USERSPACE) */ { 504 error = copyout(sp, sfsp, sizeof(*sp)); 505 free(sptmp, M_STATFS); 506 if (error != 0) { 507 vfs_unbusy(mp); 508 return (error); 509 } 510 } 511 sfsp++; 512 } 513 count++; 514 mtx_lock(&mountlist_mtx); 515 nmp = TAILQ_NEXT(mp, mnt_list); 516 vfs_unbusy(mp); 517 } 518 mtx_unlock(&mountlist_mtx); 519 if (sfsp != NULL && count > maxcount) 520 *countp = maxcount; 521 else 522 *countp = count; 523 return (0); 524 } 525 526 #ifdef COMPAT_FREEBSD4 527 /* 528 * Get old format filesystem statistics. 529 */ 530 static void freebsd4_cvtstatfs(struct statfs *, struct ostatfs *); 531 532 #ifndef _SYS_SYSPROTO_H_ 533 struct freebsd4_statfs_args { 534 char *path; 535 struct ostatfs *buf; 536 }; 537 #endif 538 int 539 freebsd4_statfs(struct thread *td, struct freebsd4_statfs_args *uap) 540 { 541 struct ostatfs osb; 542 struct statfs *sfp; 543 int error; 544 545 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 546 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp); 547 if (error == 0) { 548 freebsd4_cvtstatfs(sfp, &osb); 549 error = copyout(&osb, uap->buf, sizeof(osb)); 550 } 551 free(sfp, M_STATFS); 552 return (error); 553 } 554 555 /* 556 * Get filesystem statistics. 557 */ 558 #ifndef _SYS_SYSPROTO_H_ 559 struct freebsd4_fstatfs_args { 560 int fd; 561 struct ostatfs *buf; 562 }; 563 #endif 564 int 565 freebsd4_fstatfs(struct thread *td, struct freebsd4_fstatfs_args *uap) 566 { 567 struct ostatfs osb; 568 struct statfs *sfp; 569 int error; 570 571 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 572 error = kern_fstatfs(td, uap->fd, sfp); 573 if (error == 0) { 574 freebsd4_cvtstatfs(sfp, &osb); 575 error = copyout(&osb, uap->buf, sizeof(osb)); 576 } 577 free(sfp, M_STATFS); 578 return (error); 579 } 580 581 /* 582 * Get statistics on all filesystems. 583 */ 584 #ifndef _SYS_SYSPROTO_H_ 585 struct freebsd4_getfsstat_args { 586 struct ostatfs *buf; 587 long bufsize; 588 int mode; 589 }; 590 #endif 591 int 592 freebsd4_getfsstat(struct thread *td, struct freebsd4_getfsstat_args *uap) 593 { 594 struct statfs *buf, *sp; 595 struct ostatfs osb; 596 size_t count, size; 597 int error; 598 599 if (uap->bufsize < 0) 600 return (EINVAL); 601 count = uap->bufsize / sizeof(struct ostatfs); 602 if (count > SIZE_MAX / sizeof(struct statfs)) 603 return (EINVAL); 604 size = count * sizeof(struct statfs); 605 error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE, 606 uap->mode); 607 if (error == 0) 608 td->td_retval[0] = count; 609 if (size != 0) { 610 sp = buf; 611 while (count != 0 && error == 0) { 612 freebsd4_cvtstatfs(sp, &osb); 613 error = copyout(&osb, uap->buf, sizeof(osb)); 614 sp++; 615 uap->buf++; 616 count--; 617 } 618 free(buf, M_STATFS); 619 } 620 return (error); 621 } 622 623 /* 624 * Implement fstatfs() for (NFS) file handles. 625 */ 626 #ifndef _SYS_SYSPROTO_H_ 627 struct freebsd4_fhstatfs_args { 628 struct fhandle *u_fhp; 629 struct ostatfs *buf; 630 }; 631 #endif 632 int 633 freebsd4_fhstatfs(struct thread *td, struct freebsd4_fhstatfs_args *uap) 634 { 635 struct ostatfs osb; 636 struct statfs *sfp; 637 fhandle_t fh; 638 int error; 639 640 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t)); 641 if (error != 0) 642 return (error); 643 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 644 error = kern_fhstatfs(td, fh, sfp); 645 if (error == 0) { 646 freebsd4_cvtstatfs(sfp, &osb); 647 error = copyout(&osb, uap->buf, sizeof(osb)); 648 } 649 free(sfp, M_STATFS); 650 return (error); 651 } 652 653 /* 654 * Convert a new format statfs structure to an old format statfs structure. 655 */ 656 static void 657 freebsd4_cvtstatfs(struct statfs *nsp, struct ostatfs *osp) 658 { 659 660 statfs_scale_blocks(nsp, LONG_MAX); 661 bzero(osp, sizeof(*osp)); 662 osp->f_bsize = nsp->f_bsize; 663 osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX); 664 osp->f_blocks = nsp->f_blocks; 665 osp->f_bfree = nsp->f_bfree; 666 osp->f_bavail = nsp->f_bavail; 667 osp->f_files = MIN(nsp->f_files, LONG_MAX); 668 osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX); 669 osp->f_owner = nsp->f_owner; 670 osp->f_type = nsp->f_type; 671 osp->f_flags = nsp->f_flags; 672 osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX); 673 osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX); 674 osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX); 675 osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX); 676 strlcpy(osp->f_fstypename, nsp->f_fstypename, 677 MIN(MFSNAMELEN, OMFSNAMELEN)); 678 strlcpy(osp->f_mntonname, nsp->f_mntonname, 679 MIN(MNAMELEN, OMNAMELEN)); 680 strlcpy(osp->f_mntfromname, nsp->f_mntfromname, 681 MIN(MNAMELEN, OMNAMELEN)); 682 osp->f_fsid = nsp->f_fsid; 683 } 684 #endif /* COMPAT_FREEBSD4 */ 685 686 #if defined(COMPAT_FREEBSD11) 687 /* 688 * Get old format filesystem statistics. 689 */ 690 static void freebsd11_cvtstatfs(struct statfs *, struct freebsd11_statfs *); 691 692 int 693 freebsd11_statfs(struct thread *td, struct freebsd11_statfs_args *uap) 694 { 695 struct freebsd11_statfs osb; 696 struct statfs *sfp; 697 int error; 698 699 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 700 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp); 701 if (error == 0) { 702 freebsd11_cvtstatfs(sfp, &osb); 703 error = copyout(&osb, uap->buf, sizeof(osb)); 704 } 705 free(sfp, M_STATFS); 706 return (error); 707 } 708 709 /* 710 * Get filesystem statistics. 711 */ 712 int 713 freebsd11_fstatfs(struct thread *td, struct freebsd11_fstatfs_args *uap) 714 { 715 struct freebsd11_statfs osb; 716 struct statfs *sfp; 717 int error; 718 719 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 720 error = kern_fstatfs(td, uap->fd, sfp); 721 if (error == 0) { 722 freebsd11_cvtstatfs(sfp, &osb); 723 error = copyout(&osb, uap->buf, sizeof(osb)); 724 } 725 free(sfp, M_STATFS); 726 return (error); 727 } 728 729 /* 730 * Get statistics on all filesystems. 731 */ 732 int 733 freebsd11_getfsstat(struct thread *td, struct freebsd11_getfsstat_args *uap) 734 { 735 struct freebsd11_statfs osb; 736 struct statfs *buf, *sp; 737 size_t count, size; 738 int error; 739 740 count = uap->bufsize / sizeof(struct ostatfs); 741 size = count * sizeof(struct statfs); 742 error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE, 743 uap->mode); 744 if (error == 0) 745 td->td_retval[0] = count; 746 if (size > 0) { 747 sp = buf; 748 while (count > 0 && error == 0) { 749 freebsd11_cvtstatfs(sp, &osb); 750 error = copyout(&osb, uap->buf, sizeof(osb)); 751 sp++; 752 uap->buf++; 753 count--; 754 } 755 free(buf, M_STATFS); 756 } 757 return (error); 758 } 759 760 /* 761 * Implement fstatfs() for (NFS) file handles. 762 */ 763 int 764 freebsd11_fhstatfs(struct thread *td, struct freebsd11_fhstatfs_args *uap) 765 { 766 struct freebsd11_statfs osb; 767 struct statfs *sfp; 768 fhandle_t fh; 769 int error; 770 771 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t)); 772 if (error) 773 return (error); 774 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 775 error = kern_fhstatfs(td, fh, sfp); 776 if (error == 0) { 777 freebsd11_cvtstatfs(sfp, &osb); 778 error = copyout(&osb, uap->buf, sizeof(osb)); 779 } 780 free(sfp, M_STATFS); 781 return (error); 782 } 783 784 /* 785 * Convert a new format statfs structure to an old format statfs structure. 786 */ 787 static void 788 freebsd11_cvtstatfs(struct statfs *nsp, struct freebsd11_statfs *osp) 789 { 790 791 bzero(osp, sizeof(*osp)); 792 osp->f_version = FREEBSD11_STATFS_VERSION; 793 osp->f_type = nsp->f_type; 794 osp->f_flags = nsp->f_flags; 795 osp->f_bsize = nsp->f_bsize; 796 osp->f_iosize = nsp->f_iosize; 797 osp->f_blocks = nsp->f_blocks; 798 osp->f_bfree = nsp->f_bfree; 799 osp->f_bavail = nsp->f_bavail; 800 osp->f_files = nsp->f_files; 801 osp->f_ffree = nsp->f_ffree; 802 osp->f_syncwrites = nsp->f_syncwrites; 803 osp->f_asyncwrites = nsp->f_asyncwrites; 804 osp->f_syncreads = nsp->f_syncreads; 805 osp->f_asyncreads = nsp->f_asyncreads; 806 osp->f_namemax = nsp->f_namemax; 807 osp->f_owner = nsp->f_owner; 808 osp->f_fsid = nsp->f_fsid; 809 strlcpy(osp->f_fstypename, nsp->f_fstypename, 810 MIN(MFSNAMELEN, sizeof(osp->f_fstypename))); 811 strlcpy(osp->f_mntonname, nsp->f_mntonname, 812 MIN(MNAMELEN, sizeof(osp->f_mntonname))); 813 strlcpy(osp->f_mntfromname, nsp->f_mntfromname, 814 MIN(MNAMELEN, sizeof(osp->f_mntfromname))); 815 } 816 #endif /* COMPAT_FREEBSD11 */ 817 818 /* 819 * Change current working directory to a given file descriptor. 820 */ 821 #ifndef _SYS_SYSPROTO_H_ 822 struct fchdir_args { 823 int fd; 824 }; 825 #endif 826 int 827 sys_fchdir(struct thread *td, struct fchdir_args *uap) 828 { 829 struct vnode *vp, *tdp; 830 struct mount *mp; 831 struct file *fp; 832 cap_rights_t rights; 833 int error; 834 835 AUDIT_ARG_FD(uap->fd); 836 error = getvnode(td, uap->fd, cap_rights_init(&rights, CAP_FCHDIR), 837 &fp); 838 if (error != 0) 839 return (error); 840 vp = fp->f_vnode; 841 vrefact(vp); 842 fdrop(fp, td); 843 vn_lock(vp, LK_SHARED | LK_RETRY); 844 AUDIT_ARG_VNODE1(vp); 845 error = change_dir(vp, td); 846 while (!error && (mp = vp->v_mountedhere) != NULL) { 847 if (vfs_busy(mp, 0)) 848 continue; 849 error = VFS_ROOT(mp, LK_SHARED, &tdp); 850 vfs_unbusy(mp); 851 if (error != 0) 852 break; 853 vput(vp); 854 vp = tdp; 855 } 856 if (error != 0) { 857 vput(vp); 858 return (error); 859 } 860 VOP_UNLOCK(vp, 0); 861 pwd_chdir(td, vp); 862 return (0); 863 } 864 865 /* 866 * Change current working directory (``.''). 867 */ 868 #ifndef _SYS_SYSPROTO_H_ 869 struct chdir_args { 870 char *path; 871 }; 872 #endif 873 int 874 sys_chdir(struct thread *td, struct chdir_args *uap) 875 { 876 877 return (kern_chdir(td, uap->path, UIO_USERSPACE)); 878 } 879 880 int 881 kern_chdir(struct thread *td, char *path, enum uio_seg pathseg) 882 { 883 struct nameidata nd; 884 int error; 885 886 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1, 887 pathseg, path, td); 888 if ((error = namei(&nd)) != 0) 889 return (error); 890 if ((error = change_dir(nd.ni_vp, td)) != 0) { 891 vput(nd.ni_vp); 892 NDFREE(&nd, NDF_ONLY_PNBUF); 893 return (error); 894 } 895 VOP_UNLOCK(nd.ni_vp, 0); 896 NDFREE(&nd, NDF_ONLY_PNBUF); 897 pwd_chdir(td, nd.ni_vp); 898 return (0); 899 } 900 901 /* 902 * Change notion of root (``/'') directory. 903 */ 904 #ifndef _SYS_SYSPROTO_H_ 905 struct chroot_args { 906 char *path; 907 }; 908 #endif 909 int 910 sys_chroot(struct thread *td, struct chroot_args *uap) 911 { 912 struct nameidata nd; 913 int error; 914 915 error = priv_check(td, PRIV_VFS_CHROOT); 916 if (error != 0) 917 return (error); 918 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1, 919 UIO_USERSPACE, uap->path, td); 920 error = namei(&nd); 921 if (error != 0) 922 goto error; 923 error = change_dir(nd.ni_vp, td); 924 if (error != 0) 925 goto e_vunlock; 926 #ifdef MAC 927 error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp); 928 if (error != 0) 929 goto e_vunlock; 930 #endif 931 VOP_UNLOCK(nd.ni_vp, 0); 932 error = pwd_chroot(td, nd.ni_vp); 933 vrele(nd.ni_vp); 934 NDFREE(&nd, NDF_ONLY_PNBUF); 935 return (error); 936 e_vunlock: 937 vput(nd.ni_vp); 938 error: 939 NDFREE(&nd, NDF_ONLY_PNBUF); 940 return (error); 941 } 942 943 /* 944 * Common routine for chroot and chdir. Callers must provide a locked vnode 945 * instance. 946 */ 947 int 948 change_dir(struct vnode *vp, struct thread *td) 949 { 950 #ifdef MAC 951 int error; 952 #endif 953 954 ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked"); 955 if (vp->v_type != VDIR) 956 return (ENOTDIR); 957 #ifdef MAC 958 error = mac_vnode_check_chdir(td->td_ucred, vp); 959 if (error != 0) 960 return (error); 961 #endif 962 return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td)); 963 } 964 965 static __inline void 966 flags_to_rights(int flags, cap_rights_t *rightsp) 967 { 968 969 if (flags & O_EXEC) { 970 cap_rights_set(rightsp, CAP_FEXECVE); 971 } else { 972 switch ((flags & O_ACCMODE)) { 973 case O_RDONLY: 974 cap_rights_set(rightsp, CAP_READ); 975 break; 976 case O_RDWR: 977 cap_rights_set(rightsp, CAP_READ); 978 /* FALLTHROUGH */ 979 case O_WRONLY: 980 cap_rights_set(rightsp, CAP_WRITE); 981 if (!(flags & (O_APPEND | O_TRUNC))) 982 cap_rights_set(rightsp, CAP_SEEK); 983 break; 984 } 985 } 986 987 if (flags & O_CREAT) 988 cap_rights_set(rightsp, CAP_CREATE); 989 990 if (flags & O_TRUNC) 991 cap_rights_set(rightsp, CAP_FTRUNCATE); 992 993 if (flags & (O_SYNC | O_FSYNC)) 994 cap_rights_set(rightsp, CAP_FSYNC); 995 996 if (flags & (O_EXLOCK | O_SHLOCK)) 997 cap_rights_set(rightsp, CAP_FLOCK); 998 } 999 1000 /* 1001 * Check permissions, allocate an open file structure, and call the device 1002 * open routine if any. 1003 */ 1004 #ifndef _SYS_SYSPROTO_H_ 1005 struct open_args { 1006 char *path; 1007 int flags; 1008 int mode; 1009 }; 1010 #endif 1011 int 1012 sys_open(struct thread *td, struct open_args *uap) 1013 { 1014 1015 return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 1016 uap->flags, uap->mode)); 1017 } 1018 1019 #ifndef _SYS_SYSPROTO_H_ 1020 struct openat_args { 1021 int fd; 1022 char *path; 1023 int flag; 1024 int mode; 1025 }; 1026 #endif 1027 int 1028 sys_openat(struct thread *td, struct openat_args *uap) 1029 { 1030 1031 AUDIT_ARG_FD(uap->fd); 1032 return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag, 1033 uap->mode)); 1034 } 1035 1036 int 1037 kern_openat(struct thread *td, int fd, char *path, enum uio_seg pathseg, 1038 int flags, int mode) 1039 { 1040 struct proc *p = td->td_proc; 1041 struct filedesc *fdp = p->p_fd; 1042 struct file *fp; 1043 struct vnode *vp; 1044 struct nameidata nd; 1045 cap_rights_t rights; 1046 int cmode, error, indx; 1047 1048 indx = -1; 1049 1050 AUDIT_ARG_FFLAGS(flags); 1051 AUDIT_ARG_MODE(mode); 1052 cap_rights_init(&rights, CAP_LOOKUP); 1053 flags_to_rights(flags, &rights); 1054 /* 1055 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags 1056 * may be specified. 1057 */ 1058 if (flags & O_EXEC) { 1059 if (flags & O_ACCMODE) 1060 return (EINVAL); 1061 } else if ((flags & O_ACCMODE) == O_ACCMODE) { 1062 return (EINVAL); 1063 } else { 1064 flags = FFLAGS(flags); 1065 } 1066 1067 /* 1068 * Allocate a file structure. The descriptor to reference it 1069 * is allocated and set by finstall() below. 1070 */ 1071 error = falloc_noinstall(td, &fp); 1072 if (error != 0) 1073 return (error); 1074 /* 1075 * An extra reference on `fp' has been held for us by 1076 * falloc_noinstall(). 1077 */ 1078 /* Set the flags early so the finit in devfs can pick them up. */ 1079 fp->f_flag = flags & FMASK; 1080 cmode = ((mode & ~fdp->fd_cmask) & ALLPERMS) & ~S_ISTXT; 1081 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd, 1082 &rights, td); 1083 td->td_dupfd = -1; /* XXX check for fdopen */ 1084 error = vn_open(&nd, &flags, cmode, fp); 1085 if (error != 0) { 1086 /* 1087 * If the vn_open replaced the method vector, something 1088 * wonderous happened deep below and we just pass it up 1089 * pretending we know what we do. 1090 */ 1091 if (error == ENXIO && fp->f_ops != &badfileops) 1092 goto success; 1093 1094 /* 1095 * Handle special fdopen() case. bleh. 1096 * 1097 * Don't do this for relative (capability) lookups; we don't 1098 * understand exactly what would happen, and we don't think 1099 * that it ever should. 1100 */ 1101 if ((nd.ni_lcf & NI_LCF_STRICTRELATIVE) == 0 && 1102 (error == ENODEV || error == ENXIO) && 1103 td->td_dupfd >= 0) { 1104 error = dupfdopen(td, fdp, td->td_dupfd, flags, error, 1105 &indx); 1106 if (error == 0) 1107 goto success; 1108 } 1109 1110 goto bad; 1111 } 1112 td->td_dupfd = 0; 1113 NDFREE(&nd, NDF_ONLY_PNBUF); 1114 vp = nd.ni_vp; 1115 1116 /* 1117 * Store the vnode, for any f_type. Typically, the vnode use 1118 * count is decremented by direct call to vn_closefile() for 1119 * files that switched type in the cdevsw fdopen() method. 1120 */ 1121 fp->f_vnode = vp; 1122 /* 1123 * If the file wasn't claimed by devfs bind it to the normal 1124 * vnode operations here. 1125 */ 1126 if (fp->f_ops == &badfileops) { 1127 KASSERT(vp->v_type != VFIFO, ("Unexpected fifo.")); 1128 fp->f_seqcount = 1; 1129 finit(fp, (flags & FMASK) | (fp->f_flag & FHASLOCK), 1130 DTYPE_VNODE, vp, &vnops); 1131 } 1132 1133 VOP_UNLOCK(vp, 0); 1134 if (flags & O_TRUNC) { 1135 error = fo_truncate(fp, 0, td->td_ucred, td); 1136 if (error != 0) 1137 goto bad; 1138 } 1139 success: 1140 /* 1141 * If we haven't already installed the FD (for dupfdopen), do so now. 1142 */ 1143 if (indx == -1) { 1144 struct filecaps *fcaps; 1145 1146 #ifdef CAPABILITIES 1147 if ((nd.ni_lcf & NI_LCF_STRICTRELATIVE) != 0) 1148 fcaps = &nd.ni_filecaps; 1149 else 1150 #endif 1151 fcaps = NULL; 1152 error = finstall(td, fp, &indx, flags, fcaps); 1153 /* On success finstall() consumes fcaps. */ 1154 if (error != 0) { 1155 filecaps_free(&nd.ni_filecaps); 1156 goto bad; 1157 } 1158 } else { 1159 filecaps_free(&nd.ni_filecaps); 1160 } 1161 1162 /* 1163 * Release our private reference, leaving the one associated with 1164 * the descriptor table intact. 1165 */ 1166 fdrop(fp, td); 1167 td->td_retval[0] = indx; 1168 return (0); 1169 bad: 1170 KASSERT(indx == -1, ("indx=%d, should be -1", indx)); 1171 fdrop(fp, td); 1172 return (error); 1173 } 1174 1175 #ifdef COMPAT_43 1176 /* 1177 * Create a file. 1178 */ 1179 #ifndef _SYS_SYSPROTO_H_ 1180 struct ocreat_args { 1181 char *path; 1182 int mode; 1183 }; 1184 #endif 1185 int 1186 ocreat(struct thread *td, struct ocreat_args *uap) 1187 { 1188 1189 return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 1190 O_WRONLY | O_CREAT | O_TRUNC, uap->mode)); 1191 } 1192 #endif /* COMPAT_43 */ 1193 1194 /* 1195 * Create a special file. 1196 */ 1197 #ifndef _SYS_SYSPROTO_H_ 1198 struct mknodat_args { 1199 int fd; 1200 char *path; 1201 mode_t mode; 1202 dev_t dev; 1203 }; 1204 #endif 1205 int 1206 sys_mknodat(struct thread *td, struct mknodat_args *uap) 1207 { 1208 1209 return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode, 1210 uap->dev)); 1211 } 1212 1213 #if defined(COMPAT_FREEBSD11) 1214 int 1215 freebsd11_mknod(struct thread *td, 1216 struct freebsd11_mknod_args *uap) 1217 { 1218 1219 return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 1220 uap->mode, uap->dev)); 1221 } 1222 1223 int 1224 freebsd11_mknodat(struct thread *td, 1225 struct freebsd11_mknodat_args *uap) 1226 { 1227 1228 return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode, 1229 uap->dev)); 1230 } 1231 #endif /* COMPAT_FREEBSD11 */ 1232 1233 int 1234 kern_mknodat(struct thread *td, int fd, char *path, enum uio_seg pathseg, 1235 int mode, dev_t dev) 1236 { 1237 struct vnode *vp; 1238 struct mount *mp; 1239 struct vattr vattr; 1240 struct nameidata nd; 1241 cap_rights_t rights; 1242 int error, whiteout = 0; 1243 1244 AUDIT_ARG_MODE(mode); 1245 AUDIT_ARG_DEV(dev); 1246 switch (mode & S_IFMT) { 1247 case S_IFCHR: 1248 case S_IFBLK: 1249 error = priv_check(td, PRIV_VFS_MKNOD_DEV); 1250 if (error == 0 && dev == VNOVAL) 1251 error = EINVAL; 1252 break; 1253 case S_IFWHT: 1254 error = priv_check(td, PRIV_VFS_MKNOD_WHT); 1255 break; 1256 case S_IFIFO: 1257 if (dev == 0) 1258 return (kern_mkfifoat(td, fd, path, pathseg, mode)); 1259 /* FALLTHROUGH */ 1260 default: 1261 error = EINVAL; 1262 break; 1263 } 1264 if (error != 0) 1265 return (error); 1266 restart: 1267 bwillwrite(); 1268 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 | 1269 NOCACHE, pathseg, path, fd, cap_rights_init(&rights, CAP_MKNODAT), 1270 td); 1271 if ((error = namei(&nd)) != 0) 1272 return (error); 1273 vp = nd.ni_vp; 1274 if (vp != NULL) { 1275 NDFREE(&nd, NDF_ONLY_PNBUF); 1276 if (vp == nd.ni_dvp) 1277 vrele(nd.ni_dvp); 1278 else 1279 vput(nd.ni_dvp); 1280 vrele(vp); 1281 return (EEXIST); 1282 } else { 1283 VATTR_NULL(&vattr); 1284 vattr.va_mode = (mode & ALLPERMS) & 1285 ~td->td_proc->p_fd->fd_cmask; 1286 vattr.va_rdev = dev; 1287 whiteout = 0; 1288 1289 switch (mode & S_IFMT) { 1290 case S_IFCHR: 1291 vattr.va_type = VCHR; 1292 break; 1293 case S_IFBLK: 1294 vattr.va_type = VBLK; 1295 break; 1296 case S_IFWHT: 1297 whiteout = 1; 1298 break; 1299 default: 1300 panic("kern_mknod: invalid mode"); 1301 } 1302 } 1303 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1304 NDFREE(&nd, NDF_ONLY_PNBUF); 1305 vput(nd.ni_dvp); 1306 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0) 1307 return (error); 1308 goto restart; 1309 } 1310 #ifdef MAC 1311 if (error == 0 && !whiteout) 1312 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, 1313 &nd.ni_cnd, &vattr); 1314 #endif 1315 if (error == 0) { 1316 if (whiteout) 1317 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1318 else { 1319 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1320 &nd.ni_cnd, &vattr); 1321 if (error == 0) 1322 vput(nd.ni_vp); 1323 } 1324 } 1325 NDFREE(&nd, NDF_ONLY_PNBUF); 1326 vput(nd.ni_dvp); 1327 vn_finished_write(mp); 1328 return (error); 1329 } 1330 1331 /* 1332 * Create a named pipe. 1333 */ 1334 #ifndef _SYS_SYSPROTO_H_ 1335 struct mkfifo_args { 1336 char *path; 1337 int mode; 1338 }; 1339 #endif 1340 int 1341 sys_mkfifo(struct thread *td, struct mkfifo_args *uap) 1342 { 1343 1344 return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 1345 uap->mode)); 1346 } 1347 1348 #ifndef _SYS_SYSPROTO_H_ 1349 struct mkfifoat_args { 1350 int fd; 1351 char *path; 1352 mode_t mode; 1353 }; 1354 #endif 1355 int 1356 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap) 1357 { 1358 1359 return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE, 1360 uap->mode)); 1361 } 1362 1363 int 1364 kern_mkfifoat(struct thread *td, int fd, char *path, enum uio_seg pathseg, 1365 int mode) 1366 { 1367 struct mount *mp; 1368 struct vattr vattr; 1369 struct nameidata nd; 1370 cap_rights_t rights; 1371 int error; 1372 1373 AUDIT_ARG_MODE(mode); 1374 restart: 1375 bwillwrite(); 1376 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 | 1377 NOCACHE, pathseg, path, fd, cap_rights_init(&rights, CAP_MKFIFOAT), 1378 td); 1379 if ((error = namei(&nd)) != 0) 1380 return (error); 1381 if (nd.ni_vp != NULL) { 1382 NDFREE(&nd, NDF_ONLY_PNBUF); 1383 if (nd.ni_vp == nd.ni_dvp) 1384 vrele(nd.ni_dvp); 1385 else 1386 vput(nd.ni_dvp); 1387 vrele(nd.ni_vp); 1388 return (EEXIST); 1389 } 1390 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1391 NDFREE(&nd, NDF_ONLY_PNBUF); 1392 vput(nd.ni_dvp); 1393 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0) 1394 return (error); 1395 goto restart; 1396 } 1397 VATTR_NULL(&vattr); 1398 vattr.va_type = VFIFO; 1399 vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask; 1400 #ifdef MAC 1401 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd, 1402 &vattr); 1403 if (error != 0) 1404 goto out; 1405 #endif 1406 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1407 if (error == 0) 1408 vput(nd.ni_vp); 1409 #ifdef MAC 1410 out: 1411 #endif 1412 vput(nd.ni_dvp); 1413 vn_finished_write(mp); 1414 NDFREE(&nd, NDF_ONLY_PNBUF); 1415 return (error); 1416 } 1417 1418 /* 1419 * Make a hard file link. 1420 */ 1421 #ifndef _SYS_SYSPROTO_H_ 1422 struct link_args { 1423 char *path; 1424 char *link; 1425 }; 1426 #endif 1427 int 1428 sys_link(struct thread *td, struct link_args *uap) 1429 { 1430 1431 return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link, 1432 UIO_USERSPACE, FOLLOW)); 1433 } 1434 1435 #ifndef _SYS_SYSPROTO_H_ 1436 struct linkat_args { 1437 int fd1; 1438 char *path1; 1439 int fd2; 1440 char *path2; 1441 int flag; 1442 }; 1443 #endif 1444 int 1445 sys_linkat(struct thread *td, struct linkat_args *uap) 1446 { 1447 int flag; 1448 1449 flag = uap->flag; 1450 if (flag & ~AT_SYMLINK_FOLLOW) 1451 return (EINVAL); 1452 1453 return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2, 1454 UIO_USERSPACE, (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW)); 1455 } 1456 1457 int hardlink_check_uid = 0; 1458 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW, 1459 &hardlink_check_uid, 0, 1460 "Unprivileged processes cannot create hard links to files owned by other " 1461 "users"); 1462 static int hardlink_check_gid = 0; 1463 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW, 1464 &hardlink_check_gid, 0, 1465 "Unprivileged processes cannot create hard links to files owned by other " 1466 "groups"); 1467 1468 static int 1469 can_hardlink(struct vnode *vp, struct ucred *cred) 1470 { 1471 struct vattr va; 1472 int error; 1473 1474 if (!hardlink_check_uid && !hardlink_check_gid) 1475 return (0); 1476 1477 error = VOP_GETATTR(vp, &va, cred); 1478 if (error != 0) 1479 return (error); 1480 1481 if (hardlink_check_uid && cred->cr_uid != va.va_uid) { 1482 error = priv_check_cred(cred, PRIV_VFS_LINK, 0); 1483 if (error != 0) 1484 return (error); 1485 } 1486 1487 if (hardlink_check_gid && !groupmember(va.va_gid, cred)) { 1488 error = priv_check_cred(cred, PRIV_VFS_LINK, 0); 1489 if (error != 0) 1490 return (error); 1491 } 1492 1493 return (0); 1494 } 1495 1496 int 1497 kern_linkat(struct thread *td, int fd1, int fd2, char *path1, char *path2, 1498 enum uio_seg segflg, int follow) 1499 { 1500 struct vnode *vp; 1501 struct mount *mp; 1502 struct nameidata nd; 1503 cap_rights_t rights; 1504 int error; 1505 1506 again: 1507 bwillwrite(); 1508 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, segflg, path1, fd1, 1509 cap_rights_init(&rights, CAP_LINKAT_SOURCE), td); 1510 1511 if ((error = namei(&nd)) != 0) 1512 return (error); 1513 NDFREE(&nd, NDF_ONLY_PNBUF); 1514 vp = nd.ni_vp; 1515 if (vp->v_type == VDIR) { 1516 vrele(vp); 1517 return (EPERM); /* POSIX */ 1518 } 1519 NDINIT_ATRIGHTS(&nd, CREATE, 1520 LOCKPARENT | SAVENAME | AUDITVNODE2 | NOCACHE, segflg, path2, fd2, 1521 cap_rights_init(&rights, CAP_LINKAT_TARGET), td); 1522 if ((error = namei(&nd)) == 0) { 1523 if (nd.ni_vp != NULL) { 1524 NDFREE(&nd, NDF_ONLY_PNBUF); 1525 if (nd.ni_dvp == nd.ni_vp) 1526 vrele(nd.ni_dvp); 1527 else 1528 vput(nd.ni_dvp); 1529 vrele(nd.ni_vp); 1530 vrele(vp); 1531 return (EEXIST); 1532 } else if (nd.ni_dvp->v_mount != vp->v_mount) { 1533 /* 1534 * Cross-device link. No need to recheck 1535 * vp->v_type, since it cannot change, except 1536 * to VBAD. 1537 */ 1538 NDFREE(&nd, NDF_ONLY_PNBUF); 1539 vput(nd.ni_dvp); 1540 vrele(vp); 1541 return (EXDEV); 1542 } else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) { 1543 error = can_hardlink(vp, td->td_ucred); 1544 #ifdef MAC 1545 if (error == 0) 1546 error = mac_vnode_check_link(td->td_ucred, 1547 nd.ni_dvp, vp, &nd.ni_cnd); 1548 #endif 1549 if (error != 0) { 1550 vput(vp); 1551 vput(nd.ni_dvp); 1552 NDFREE(&nd, NDF_ONLY_PNBUF); 1553 return (error); 1554 } 1555 error = vn_start_write(vp, &mp, V_NOWAIT); 1556 if (error != 0) { 1557 vput(vp); 1558 vput(nd.ni_dvp); 1559 NDFREE(&nd, NDF_ONLY_PNBUF); 1560 error = vn_start_write(NULL, &mp, 1561 V_XSLEEP | PCATCH); 1562 if (error != 0) 1563 return (error); 1564 goto again; 1565 } 1566 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1567 VOP_UNLOCK(vp, 0); 1568 vput(nd.ni_dvp); 1569 vn_finished_write(mp); 1570 NDFREE(&nd, NDF_ONLY_PNBUF); 1571 } else { 1572 vput(nd.ni_dvp); 1573 NDFREE(&nd, NDF_ONLY_PNBUF); 1574 vrele(vp); 1575 goto again; 1576 } 1577 } 1578 vrele(vp); 1579 return (error); 1580 } 1581 1582 /* 1583 * Make a symbolic link. 1584 */ 1585 #ifndef _SYS_SYSPROTO_H_ 1586 struct symlink_args { 1587 char *path; 1588 char *link; 1589 }; 1590 #endif 1591 int 1592 sys_symlink(struct thread *td, struct symlink_args *uap) 1593 { 1594 1595 return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link, 1596 UIO_USERSPACE)); 1597 } 1598 1599 #ifndef _SYS_SYSPROTO_H_ 1600 struct symlinkat_args { 1601 char *path; 1602 int fd; 1603 char *path2; 1604 }; 1605 #endif 1606 int 1607 sys_symlinkat(struct thread *td, struct symlinkat_args *uap) 1608 { 1609 1610 return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2, 1611 UIO_USERSPACE)); 1612 } 1613 1614 int 1615 kern_symlinkat(struct thread *td, char *path1, int fd, char *path2, 1616 enum uio_seg segflg) 1617 { 1618 struct mount *mp; 1619 struct vattr vattr; 1620 char *syspath; 1621 struct nameidata nd; 1622 int error; 1623 cap_rights_t rights; 1624 1625 if (segflg == UIO_SYSSPACE) { 1626 syspath = path1; 1627 } else { 1628 syspath = uma_zalloc(namei_zone, M_WAITOK); 1629 if ((error = copyinstr(path1, syspath, MAXPATHLEN, NULL)) != 0) 1630 goto out; 1631 } 1632 AUDIT_ARG_TEXT(syspath); 1633 restart: 1634 bwillwrite(); 1635 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 | 1636 NOCACHE, segflg, path2, fd, cap_rights_init(&rights, CAP_SYMLINKAT), 1637 td); 1638 if ((error = namei(&nd)) != 0) 1639 goto out; 1640 if (nd.ni_vp) { 1641 NDFREE(&nd, NDF_ONLY_PNBUF); 1642 if (nd.ni_vp == nd.ni_dvp) 1643 vrele(nd.ni_dvp); 1644 else 1645 vput(nd.ni_dvp); 1646 vrele(nd.ni_vp); 1647 error = EEXIST; 1648 goto out; 1649 } 1650 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1651 NDFREE(&nd, NDF_ONLY_PNBUF); 1652 vput(nd.ni_dvp); 1653 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0) 1654 goto out; 1655 goto restart; 1656 } 1657 VATTR_NULL(&vattr); 1658 vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask; 1659 #ifdef MAC 1660 vattr.va_type = VLNK; 1661 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd, 1662 &vattr); 1663 if (error != 0) 1664 goto out2; 1665 #endif 1666 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath); 1667 if (error == 0) 1668 vput(nd.ni_vp); 1669 #ifdef MAC 1670 out2: 1671 #endif 1672 NDFREE(&nd, NDF_ONLY_PNBUF); 1673 vput(nd.ni_dvp); 1674 vn_finished_write(mp); 1675 out: 1676 if (segflg != UIO_SYSSPACE) 1677 uma_zfree(namei_zone, syspath); 1678 return (error); 1679 } 1680 1681 /* 1682 * Delete a whiteout from the filesystem. 1683 */ 1684 #ifndef _SYS_SYSPROTO_H_ 1685 struct undelete_args { 1686 char *path; 1687 }; 1688 #endif 1689 int 1690 sys_undelete(struct thread *td, struct undelete_args *uap) 1691 { 1692 struct mount *mp; 1693 struct nameidata nd; 1694 int error; 1695 1696 restart: 1697 bwillwrite(); 1698 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1, 1699 UIO_USERSPACE, uap->path, td); 1700 error = namei(&nd); 1701 if (error != 0) 1702 return (error); 1703 1704 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 1705 NDFREE(&nd, NDF_ONLY_PNBUF); 1706 if (nd.ni_vp == nd.ni_dvp) 1707 vrele(nd.ni_dvp); 1708 else 1709 vput(nd.ni_dvp); 1710 if (nd.ni_vp) 1711 vrele(nd.ni_vp); 1712 return (EEXIST); 1713 } 1714 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1715 NDFREE(&nd, NDF_ONLY_PNBUF); 1716 vput(nd.ni_dvp); 1717 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0) 1718 return (error); 1719 goto restart; 1720 } 1721 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE); 1722 NDFREE(&nd, NDF_ONLY_PNBUF); 1723 vput(nd.ni_dvp); 1724 vn_finished_write(mp); 1725 return (error); 1726 } 1727 1728 /* 1729 * Delete a name from the filesystem. 1730 */ 1731 #ifndef _SYS_SYSPROTO_H_ 1732 struct unlink_args { 1733 char *path; 1734 }; 1735 #endif 1736 int 1737 sys_unlink(struct thread *td, struct unlink_args *uap) 1738 { 1739 1740 return (kern_unlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 0)); 1741 } 1742 1743 #ifndef _SYS_SYSPROTO_H_ 1744 struct unlinkat_args { 1745 int fd; 1746 char *path; 1747 int flag; 1748 }; 1749 #endif 1750 int 1751 sys_unlinkat(struct thread *td, struct unlinkat_args *uap) 1752 { 1753 int flag = uap->flag; 1754 int fd = uap->fd; 1755 char *path = uap->path; 1756 1757 if (flag & ~AT_REMOVEDIR) 1758 return (EINVAL); 1759 1760 if (flag & AT_REMOVEDIR) 1761 return (kern_rmdirat(td, fd, path, UIO_USERSPACE)); 1762 else 1763 return (kern_unlinkat(td, fd, path, UIO_USERSPACE, 0)); 1764 } 1765 1766 int 1767 kern_unlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg, 1768 ino_t oldinum) 1769 { 1770 struct mount *mp; 1771 struct vnode *vp; 1772 struct nameidata nd; 1773 struct stat sb; 1774 cap_rights_t rights; 1775 int error; 1776 1777 restart: 1778 bwillwrite(); 1779 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1, 1780 pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td); 1781 if ((error = namei(&nd)) != 0) 1782 return (error == EINVAL ? EPERM : error); 1783 vp = nd.ni_vp; 1784 if (vp->v_type == VDIR && oldinum == 0) { 1785 error = EPERM; /* POSIX */ 1786 } else if (oldinum != 0 && 1787 ((error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td)) == 0) && 1788 sb.st_ino != oldinum) { 1789 error = EIDRM; /* Identifier removed */ 1790 } else { 1791 /* 1792 * The root of a mounted filesystem cannot be deleted. 1793 * 1794 * XXX: can this only be a VDIR case? 1795 */ 1796 if (vp->v_vflag & VV_ROOT) 1797 error = EBUSY; 1798 } 1799 if (error == 0) { 1800 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1801 NDFREE(&nd, NDF_ONLY_PNBUF); 1802 vput(nd.ni_dvp); 1803 if (vp == nd.ni_dvp) 1804 vrele(vp); 1805 else 1806 vput(vp); 1807 if ((error = vn_start_write(NULL, &mp, 1808 V_XSLEEP | PCATCH)) != 0) 1809 return (error); 1810 goto restart; 1811 } 1812 #ifdef MAC 1813 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp, 1814 &nd.ni_cnd); 1815 if (error != 0) 1816 goto out; 1817 #endif 1818 vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK); 1819 error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd); 1820 #ifdef MAC 1821 out: 1822 #endif 1823 vn_finished_write(mp); 1824 } 1825 NDFREE(&nd, NDF_ONLY_PNBUF); 1826 vput(nd.ni_dvp); 1827 if (vp == nd.ni_dvp) 1828 vrele(vp); 1829 else 1830 vput(vp); 1831 return (error); 1832 } 1833 1834 /* 1835 * Reposition read/write file offset. 1836 */ 1837 #ifndef _SYS_SYSPROTO_H_ 1838 struct lseek_args { 1839 int fd; 1840 int pad; 1841 off_t offset; 1842 int whence; 1843 }; 1844 #endif 1845 int 1846 sys_lseek(struct thread *td, struct lseek_args *uap) 1847 { 1848 1849 return (kern_lseek(td, uap->fd, uap->offset, uap->whence)); 1850 } 1851 1852 int 1853 kern_lseek(struct thread *td, int fd, off_t offset, int whence) 1854 { 1855 struct file *fp; 1856 cap_rights_t rights; 1857 int error; 1858 1859 AUDIT_ARG_FD(fd); 1860 error = fget(td, fd, cap_rights_init(&rights, CAP_SEEK), &fp); 1861 if (error != 0) 1862 return (error); 1863 error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ? 1864 fo_seek(fp, offset, whence, td) : ESPIPE; 1865 fdrop(fp, td); 1866 return (error); 1867 } 1868 1869 #if defined(COMPAT_43) 1870 /* 1871 * Reposition read/write file offset. 1872 */ 1873 #ifndef _SYS_SYSPROTO_H_ 1874 struct olseek_args { 1875 int fd; 1876 long offset; 1877 int whence; 1878 }; 1879 #endif 1880 int 1881 olseek(struct thread *td, struct olseek_args *uap) 1882 { 1883 1884 return (kern_lseek(td, uap->fd, uap->offset, uap->whence)); 1885 } 1886 #endif /* COMPAT_43 */ 1887 1888 #if defined(COMPAT_FREEBSD6) 1889 /* Version with the 'pad' argument */ 1890 int 1891 freebsd6_lseek(struct thread *td, struct freebsd6_lseek_args *uap) 1892 { 1893 1894 return (kern_lseek(td, uap->fd, uap->offset, uap->whence)); 1895 } 1896 #endif 1897 1898 /* 1899 * Check access permissions using passed credentials. 1900 */ 1901 static int 1902 vn_access(struct vnode *vp, int user_flags, struct ucred *cred, 1903 struct thread *td) 1904 { 1905 accmode_t accmode; 1906 int error; 1907 1908 /* Flags == 0 means only check for existence. */ 1909 if (user_flags == 0) 1910 return (0); 1911 1912 accmode = 0; 1913 if (user_flags & R_OK) 1914 accmode |= VREAD; 1915 if (user_flags & W_OK) 1916 accmode |= VWRITE; 1917 if (user_flags & X_OK) 1918 accmode |= VEXEC; 1919 #ifdef MAC 1920 error = mac_vnode_check_access(cred, vp, accmode); 1921 if (error != 0) 1922 return (error); 1923 #endif 1924 if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0) 1925 error = VOP_ACCESS(vp, accmode, cred, td); 1926 return (error); 1927 } 1928 1929 /* 1930 * Check access permissions using "real" credentials. 1931 */ 1932 #ifndef _SYS_SYSPROTO_H_ 1933 struct access_args { 1934 char *path; 1935 int amode; 1936 }; 1937 #endif 1938 int 1939 sys_access(struct thread *td, struct access_args *uap) 1940 { 1941 1942 return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 1943 0, uap->amode)); 1944 } 1945 1946 #ifndef _SYS_SYSPROTO_H_ 1947 struct faccessat_args { 1948 int dirfd; 1949 char *path; 1950 int amode; 1951 int flag; 1952 } 1953 #endif 1954 int 1955 sys_faccessat(struct thread *td, struct faccessat_args *uap) 1956 { 1957 1958 return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag, 1959 uap->amode)); 1960 } 1961 1962 int 1963 kern_accessat(struct thread *td, int fd, char *path, enum uio_seg pathseg, 1964 int flag, int amode) 1965 { 1966 struct ucred *cred, *usecred; 1967 struct vnode *vp; 1968 struct nameidata nd; 1969 cap_rights_t rights; 1970 int error; 1971 1972 if (flag & ~AT_EACCESS) 1973 return (EINVAL); 1974 if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0) 1975 return (EINVAL); 1976 1977 /* 1978 * Create and modify a temporary credential instead of one that 1979 * is potentially shared (if we need one). 1980 */ 1981 cred = td->td_ucred; 1982 if ((flag & AT_EACCESS) == 0 && 1983 ((cred->cr_uid != cred->cr_ruid || 1984 cred->cr_rgid != cred->cr_groups[0]))) { 1985 usecred = crdup(cred); 1986 usecred->cr_uid = cred->cr_ruid; 1987 usecred->cr_groups[0] = cred->cr_rgid; 1988 td->td_ucred = usecred; 1989 } else 1990 usecred = cred; 1991 AUDIT_ARG_VALUE(amode); 1992 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | 1993 AUDITVNODE1, pathseg, path, fd, cap_rights_init(&rights, CAP_FSTAT), 1994 td); 1995 if ((error = namei(&nd)) != 0) 1996 goto out; 1997 vp = nd.ni_vp; 1998 1999 error = vn_access(vp, amode, usecred, td); 2000 NDFREE(&nd, NDF_ONLY_PNBUF); 2001 vput(vp); 2002 out: 2003 if (usecred != cred) { 2004 td->td_ucred = cred; 2005 crfree(usecred); 2006 } 2007 return (error); 2008 } 2009 2010 /* 2011 * Check access permissions using "effective" credentials. 2012 */ 2013 #ifndef _SYS_SYSPROTO_H_ 2014 struct eaccess_args { 2015 char *path; 2016 int amode; 2017 }; 2018 #endif 2019 int 2020 sys_eaccess(struct thread *td, struct eaccess_args *uap) 2021 { 2022 2023 return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2024 AT_EACCESS, uap->amode)); 2025 } 2026 2027 #if defined(COMPAT_43) 2028 /* 2029 * Get file status; this version follows links. 2030 */ 2031 #ifndef _SYS_SYSPROTO_H_ 2032 struct ostat_args { 2033 char *path; 2034 struct ostat *ub; 2035 }; 2036 #endif 2037 int 2038 ostat(struct thread *td, struct ostat_args *uap) 2039 { 2040 struct stat sb; 2041 struct ostat osb; 2042 int error; 2043 2044 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE, 2045 &sb, NULL); 2046 if (error != 0) 2047 return (error); 2048 cvtstat(&sb, &osb); 2049 return (copyout(&osb, uap->ub, sizeof (osb))); 2050 } 2051 2052 /* 2053 * Get file status; this version does not follow links. 2054 */ 2055 #ifndef _SYS_SYSPROTO_H_ 2056 struct olstat_args { 2057 char *path; 2058 struct ostat *ub; 2059 }; 2060 #endif 2061 int 2062 olstat(struct thread *td, struct olstat_args *uap) 2063 { 2064 struct stat sb; 2065 struct ostat osb; 2066 int error; 2067 2068 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path, 2069 UIO_USERSPACE, &sb, NULL); 2070 if (error != 0) 2071 return (error); 2072 cvtstat(&sb, &osb); 2073 return (copyout(&osb, uap->ub, sizeof (osb))); 2074 } 2075 2076 /* 2077 * Convert from an old to a new stat structure. 2078 */ 2079 void 2080 cvtstat(struct stat *st, struct ostat *ost) 2081 { 2082 2083 bzero(ost, sizeof(*ost)); 2084 ost->st_dev = st->st_dev; 2085 ost->st_ino = st->st_ino; 2086 ost->st_mode = st->st_mode; 2087 ost->st_nlink = st->st_nlink; 2088 ost->st_uid = st->st_uid; 2089 ost->st_gid = st->st_gid; 2090 ost->st_rdev = st->st_rdev; 2091 if (st->st_size < (quad_t)1 << 32) 2092 ost->st_size = st->st_size; 2093 else 2094 ost->st_size = -2; 2095 ost->st_atim = st->st_atim; 2096 ost->st_mtim = st->st_mtim; 2097 ost->st_ctim = st->st_ctim; 2098 ost->st_blksize = st->st_blksize; 2099 ost->st_blocks = st->st_blocks; 2100 ost->st_flags = st->st_flags; 2101 ost->st_gen = st->st_gen; 2102 } 2103 #endif /* COMPAT_43 */ 2104 2105 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11) 2106 int ino64_trunc_error; 2107 SYSCTL_INT(_vfs, OID_AUTO, ino64_trunc_error, CTLFLAG_RW, 2108 &ino64_trunc_error, 0, 2109 "Error on truncation of inode number, device id or link count"); 2110 int 2111 freebsd11_cvtstat(struct stat *st, struct freebsd11_stat *ost) 2112 { 2113 2114 ost->st_dev = st->st_dev; 2115 ost->st_ino = st->st_ino; 2116 if (ost->st_ino != st->st_ino) { 2117 switch (ino64_trunc_error) { 2118 default: 2119 case 0: 2120 break; 2121 case 1: 2122 return (EOVERFLOW); 2123 case 2: 2124 ost->st_ino = UINT32_MAX; 2125 break; 2126 } 2127 } 2128 ost->st_mode = st->st_mode; 2129 ost->st_nlink = st->st_nlink; 2130 if (ost->st_nlink != st->st_nlink) { 2131 switch (ino64_trunc_error) { 2132 default: 2133 case 0: 2134 break; 2135 case 1: 2136 return (EOVERFLOW); 2137 case 2: 2138 ost->st_nlink = UINT16_MAX; 2139 break; 2140 } 2141 } 2142 ost->st_uid = st->st_uid; 2143 ost->st_gid = st->st_gid; 2144 ost->st_rdev = st->st_rdev; 2145 ost->st_atim = st->st_atim; 2146 ost->st_mtim = st->st_mtim; 2147 ost->st_ctim = st->st_ctim; 2148 ost->st_size = st->st_size; 2149 ost->st_blocks = st->st_blocks; 2150 ost->st_blksize = st->st_blksize; 2151 ost->st_flags = st->st_flags; 2152 ost->st_gen = st->st_gen; 2153 ost->st_lspare = 0; 2154 ost->st_birthtim = st->st_birthtim; 2155 bzero((char *)&ost->st_birthtim + sizeof(ost->st_birthtim), 2156 sizeof(*ost) - offsetof(struct freebsd11_stat, 2157 st_birthtim) - sizeof(ost->st_birthtim)); 2158 return (0); 2159 } 2160 2161 int 2162 freebsd11_stat(struct thread *td, struct freebsd11_stat_args* uap) 2163 { 2164 struct stat sb; 2165 struct freebsd11_stat osb; 2166 int error; 2167 2168 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE, 2169 &sb, NULL); 2170 if (error != 0) 2171 return (error); 2172 error = freebsd11_cvtstat(&sb, &osb); 2173 if (error == 0) 2174 error = copyout(&osb, uap->ub, sizeof(osb)); 2175 return (error); 2176 } 2177 2178 int 2179 freebsd11_lstat(struct thread *td, struct freebsd11_lstat_args* uap) 2180 { 2181 struct stat sb; 2182 struct freebsd11_stat osb; 2183 int error; 2184 2185 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path, 2186 UIO_USERSPACE, &sb, NULL); 2187 if (error != 0) 2188 return (error); 2189 error = freebsd11_cvtstat(&sb, &osb); 2190 if (error == 0) 2191 error = copyout(&osb, uap->ub, sizeof(osb)); 2192 return (error); 2193 } 2194 2195 int 2196 freebsd11_fhstat(struct thread *td, struct freebsd11_fhstat_args* uap) 2197 { 2198 struct fhandle fh; 2199 struct stat sb; 2200 struct freebsd11_stat osb; 2201 int error; 2202 2203 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t)); 2204 if (error != 0) 2205 return (error); 2206 error = kern_fhstat(td, fh, &sb); 2207 if (error != 0) 2208 return (error); 2209 error = freebsd11_cvtstat(&sb, &osb); 2210 if (error == 0) 2211 error = copyout(&osb, uap->sb, sizeof(osb)); 2212 return (error); 2213 } 2214 2215 int 2216 freebsd11_fstatat(struct thread *td, struct freebsd11_fstatat_args* uap) 2217 { 2218 struct stat sb; 2219 struct freebsd11_stat osb; 2220 int error; 2221 2222 error = kern_statat(td, uap->flag, uap->fd, uap->path, 2223 UIO_USERSPACE, &sb, NULL); 2224 if (error != 0) 2225 return (error); 2226 error = freebsd11_cvtstat(&sb, &osb); 2227 if (error == 0) 2228 error = copyout(&osb, uap->buf, sizeof(osb)); 2229 return (error); 2230 } 2231 #endif /* COMPAT_FREEBSD11 */ 2232 2233 /* 2234 * Get file status 2235 */ 2236 #ifndef _SYS_SYSPROTO_H_ 2237 struct fstatat_args { 2238 int fd; 2239 char *path; 2240 struct stat *buf; 2241 int flag; 2242 } 2243 #endif 2244 int 2245 sys_fstatat(struct thread *td, struct fstatat_args *uap) 2246 { 2247 struct stat sb; 2248 int error; 2249 2250 error = kern_statat(td, uap->flag, uap->fd, uap->path, 2251 UIO_USERSPACE, &sb, NULL); 2252 if (error == 0) 2253 error = copyout(&sb, uap->buf, sizeof (sb)); 2254 return (error); 2255 } 2256 2257 int 2258 kern_statat(struct thread *td, int flag, int fd, char *path, 2259 enum uio_seg pathseg, struct stat *sbp, 2260 void (*hook)(struct vnode *vp, struct stat *sbp)) 2261 { 2262 struct nameidata nd; 2263 struct stat sb; 2264 cap_rights_t rights; 2265 int error; 2266 2267 if (flag & ~AT_SYMLINK_NOFOLLOW) 2268 return (EINVAL); 2269 2270 NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : 2271 FOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1, pathseg, path, fd, 2272 cap_rights_init(&rights, CAP_FSTAT), td); 2273 2274 if ((error = namei(&nd)) != 0) 2275 return (error); 2276 error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td); 2277 if (error == 0) { 2278 SDT_PROBE2(vfs, , stat, mode, path, sb.st_mode); 2279 if (S_ISREG(sb.st_mode)) 2280 SDT_PROBE2(vfs, , stat, reg, path, pathseg); 2281 if (__predict_false(hook != NULL)) 2282 hook(nd.ni_vp, &sb); 2283 } 2284 NDFREE(&nd, NDF_ONLY_PNBUF); 2285 vput(nd.ni_vp); 2286 if (error != 0) 2287 return (error); 2288 #ifdef __STAT_TIME_T_EXT 2289 sb.st_atim_ext = 0; 2290 sb.st_mtim_ext = 0; 2291 sb.st_ctim_ext = 0; 2292 sb.st_btim_ext = 0; 2293 #endif 2294 *sbp = sb; 2295 #ifdef KTRACE 2296 if (KTRPOINT(td, KTR_STRUCT)) 2297 ktrstat(&sb); 2298 #endif 2299 return (0); 2300 } 2301 2302 #if defined(COMPAT_FREEBSD11) 2303 /* 2304 * Implementation of the NetBSD [l]stat() functions. 2305 */ 2306 void 2307 freebsd11_cvtnstat(struct stat *sb, struct nstat *nsb) 2308 { 2309 2310 bzero(nsb, sizeof(*nsb)); 2311 nsb->st_dev = sb->st_dev; 2312 nsb->st_ino = sb->st_ino; 2313 nsb->st_mode = sb->st_mode; 2314 nsb->st_nlink = sb->st_nlink; 2315 nsb->st_uid = sb->st_uid; 2316 nsb->st_gid = sb->st_gid; 2317 nsb->st_rdev = sb->st_rdev; 2318 nsb->st_atim = sb->st_atim; 2319 nsb->st_mtim = sb->st_mtim; 2320 nsb->st_ctim = sb->st_ctim; 2321 nsb->st_size = sb->st_size; 2322 nsb->st_blocks = sb->st_blocks; 2323 nsb->st_blksize = sb->st_blksize; 2324 nsb->st_flags = sb->st_flags; 2325 nsb->st_gen = sb->st_gen; 2326 nsb->st_birthtim = sb->st_birthtim; 2327 } 2328 2329 #ifndef _SYS_SYSPROTO_H_ 2330 struct freebsd11_nstat_args { 2331 char *path; 2332 struct nstat *ub; 2333 }; 2334 #endif 2335 int 2336 freebsd11_nstat(struct thread *td, struct freebsd11_nstat_args *uap) 2337 { 2338 struct stat sb; 2339 struct nstat nsb; 2340 int error; 2341 2342 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE, 2343 &sb, NULL); 2344 if (error != 0) 2345 return (error); 2346 freebsd11_cvtnstat(&sb, &nsb); 2347 return (copyout(&nsb, uap->ub, sizeof (nsb))); 2348 } 2349 2350 /* 2351 * NetBSD lstat. Get file status; this version does not follow links. 2352 */ 2353 #ifndef _SYS_SYSPROTO_H_ 2354 struct freebsd11_nlstat_args { 2355 char *path; 2356 struct nstat *ub; 2357 }; 2358 #endif 2359 int 2360 freebsd11_nlstat(struct thread *td, struct freebsd11_nlstat_args *uap) 2361 { 2362 struct stat sb; 2363 struct nstat nsb; 2364 int error; 2365 2366 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path, 2367 UIO_USERSPACE, &sb, NULL); 2368 if (error != 0) 2369 return (error); 2370 freebsd11_cvtnstat(&sb, &nsb); 2371 return (copyout(&nsb, uap->ub, sizeof (nsb))); 2372 } 2373 #endif /* COMPAT_FREEBSD11 */ 2374 2375 /* 2376 * Get configurable pathname variables. 2377 */ 2378 #ifndef _SYS_SYSPROTO_H_ 2379 struct pathconf_args { 2380 char *path; 2381 int name; 2382 }; 2383 #endif 2384 int 2385 sys_pathconf(struct thread *td, struct pathconf_args *uap) 2386 { 2387 2388 return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW)); 2389 } 2390 2391 #ifndef _SYS_SYSPROTO_H_ 2392 struct lpathconf_args { 2393 char *path; 2394 int name; 2395 }; 2396 #endif 2397 int 2398 sys_lpathconf(struct thread *td, struct lpathconf_args *uap) 2399 { 2400 2401 return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, 2402 NOFOLLOW)); 2403 } 2404 2405 int 2406 kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name, 2407 u_long flags) 2408 { 2409 struct nameidata nd; 2410 int error; 2411 2412 NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags, 2413 pathseg, path, td); 2414 if ((error = namei(&nd)) != 0) 2415 return (error); 2416 NDFREE(&nd, NDF_ONLY_PNBUF); 2417 2418 error = VOP_PATHCONF(nd.ni_vp, name, td->td_retval); 2419 vput(nd.ni_vp); 2420 return (error); 2421 } 2422 2423 /* 2424 * Return target name of a symbolic link. 2425 */ 2426 #ifndef _SYS_SYSPROTO_H_ 2427 struct readlink_args { 2428 char *path; 2429 char *buf; 2430 size_t count; 2431 }; 2432 #endif 2433 int 2434 sys_readlink(struct thread *td, struct readlink_args *uap) 2435 { 2436 2437 return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2438 uap->buf, UIO_USERSPACE, uap->count)); 2439 } 2440 #ifndef _SYS_SYSPROTO_H_ 2441 struct readlinkat_args { 2442 int fd; 2443 char *path; 2444 char *buf; 2445 size_t bufsize; 2446 }; 2447 #endif 2448 int 2449 sys_readlinkat(struct thread *td, struct readlinkat_args *uap) 2450 { 2451 2452 return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE, 2453 uap->buf, UIO_USERSPACE, uap->bufsize)); 2454 } 2455 2456 int 2457 kern_readlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg, 2458 char *buf, enum uio_seg bufseg, size_t count) 2459 { 2460 struct vnode *vp; 2461 struct iovec aiov; 2462 struct uio auio; 2463 struct nameidata nd; 2464 int error; 2465 2466 if (count > IOSIZE_MAX) 2467 return (EINVAL); 2468 2469 NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1, 2470 pathseg, path, fd, td); 2471 2472 if ((error = namei(&nd)) != 0) 2473 return (error); 2474 NDFREE(&nd, NDF_ONLY_PNBUF); 2475 vp = nd.ni_vp; 2476 #ifdef MAC 2477 error = mac_vnode_check_readlink(td->td_ucred, vp); 2478 if (error != 0) { 2479 vput(vp); 2480 return (error); 2481 } 2482 #endif 2483 if (vp->v_type != VLNK && (vp->v_vflag & VV_READLINK) == 0) 2484 error = EINVAL; 2485 else { 2486 aiov.iov_base = buf; 2487 aiov.iov_len = count; 2488 auio.uio_iov = &aiov; 2489 auio.uio_iovcnt = 1; 2490 auio.uio_offset = 0; 2491 auio.uio_rw = UIO_READ; 2492 auio.uio_segflg = bufseg; 2493 auio.uio_td = td; 2494 auio.uio_resid = count; 2495 error = VOP_READLINK(vp, &auio, td->td_ucred); 2496 td->td_retval[0] = count - auio.uio_resid; 2497 } 2498 vput(vp); 2499 return (error); 2500 } 2501 2502 /* 2503 * Common implementation code for chflags() and fchflags(). 2504 */ 2505 static int 2506 setfflags(struct thread *td, struct vnode *vp, u_long flags) 2507 { 2508 struct mount *mp; 2509 struct vattr vattr; 2510 int error; 2511 2512 /* We can't support the value matching VNOVAL. */ 2513 if (flags == VNOVAL) 2514 return (EOPNOTSUPP); 2515 2516 /* 2517 * Prevent non-root users from setting flags on devices. When 2518 * a device is reused, users can retain ownership of the device 2519 * if they are allowed to set flags and programs assume that 2520 * chown can't fail when done as root. 2521 */ 2522 if (vp->v_type == VCHR || vp->v_type == VBLK) { 2523 error = priv_check(td, PRIV_VFS_CHFLAGS_DEV); 2524 if (error != 0) 2525 return (error); 2526 } 2527 2528 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) 2529 return (error); 2530 VATTR_NULL(&vattr); 2531 vattr.va_flags = flags; 2532 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2533 #ifdef MAC 2534 error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags); 2535 if (error == 0) 2536 #endif 2537 error = VOP_SETATTR(vp, &vattr, td->td_ucred); 2538 VOP_UNLOCK(vp, 0); 2539 vn_finished_write(mp); 2540 return (error); 2541 } 2542 2543 /* 2544 * Change flags of a file given a path name. 2545 */ 2546 #ifndef _SYS_SYSPROTO_H_ 2547 struct chflags_args { 2548 const char *path; 2549 u_long flags; 2550 }; 2551 #endif 2552 int 2553 sys_chflags(struct thread *td, struct chflags_args *uap) 2554 { 2555 2556 return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2557 uap->flags, 0)); 2558 } 2559 2560 #ifndef _SYS_SYSPROTO_H_ 2561 struct chflagsat_args { 2562 int fd; 2563 const char *path; 2564 u_long flags; 2565 int atflag; 2566 } 2567 #endif 2568 int 2569 sys_chflagsat(struct thread *td, struct chflagsat_args *uap) 2570 { 2571 int fd = uap->fd; 2572 const char *path = uap->path; 2573 u_long flags = uap->flags; 2574 int atflag = uap->atflag; 2575 2576 if (atflag & ~AT_SYMLINK_NOFOLLOW) 2577 return (EINVAL); 2578 2579 return (kern_chflagsat(td, fd, path, UIO_USERSPACE, flags, atflag)); 2580 } 2581 2582 /* 2583 * Same as chflags() but doesn't follow symlinks. 2584 */ 2585 #ifndef _SYS_SYSPROTO_H_ 2586 struct lchflags_args { 2587 const char *path; 2588 u_long flags; 2589 }; 2590 #endif 2591 int 2592 sys_lchflags(struct thread *td, struct lchflags_args *uap) 2593 { 2594 2595 return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2596 uap->flags, AT_SYMLINK_NOFOLLOW)); 2597 } 2598 2599 static int 2600 kern_chflagsat(struct thread *td, int fd, const char *path, 2601 enum uio_seg pathseg, u_long flags, int atflag) 2602 { 2603 struct nameidata nd; 2604 cap_rights_t rights; 2605 int error, follow; 2606 2607 AUDIT_ARG_FFLAGS(flags); 2608 follow = (atflag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW; 2609 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd, 2610 cap_rights_init(&rights, CAP_FCHFLAGS), td); 2611 if ((error = namei(&nd)) != 0) 2612 return (error); 2613 NDFREE(&nd, NDF_ONLY_PNBUF); 2614 error = setfflags(td, nd.ni_vp, flags); 2615 vrele(nd.ni_vp); 2616 return (error); 2617 } 2618 2619 /* 2620 * Change flags of a file given a file descriptor. 2621 */ 2622 #ifndef _SYS_SYSPROTO_H_ 2623 struct fchflags_args { 2624 int fd; 2625 u_long flags; 2626 }; 2627 #endif 2628 int 2629 sys_fchflags(struct thread *td, struct fchflags_args *uap) 2630 { 2631 struct file *fp; 2632 cap_rights_t rights; 2633 int error; 2634 2635 AUDIT_ARG_FD(uap->fd); 2636 AUDIT_ARG_FFLAGS(uap->flags); 2637 error = getvnode(td, uap->fd, cap_rights_init(&rights, CAP_FCHFLAGS), 2638 &fp); 2639 if (error != 0) 2640 return (error); 2641 #ifdef AUDIT 2642 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY); 2643 AUDIT_ARG_VNODE1(fp->f_vnode); 2644 VOP_UNLOCK(fp->f_vnode, 0); 2645 #endif 2646 error = setfflags(td, fp->f_vnode, uap->flags); 2647 fdrop(fp, td); 2648 return (error); 2649 } 2650 2651 /* 2652 * Common implementation code for chmod(), lchmod() and fchmod(). 2653 */ 2654 int 2655 setfmode(struct thread *td, struct ucred *cred, struct vnode *vp, int mode) 2656 { 2657 struct mount *mp; 2658 struct vattr vattr; 2659 int error; 2660 2661 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) 2662 return (error); 2663 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2664 VATTR_NULL(&vattr); 2665 vattr.va_mode = mode & ALLPERMS; 2666 #ifdef MAC 2667 error = mac_vnode_check_setmode(cred, vp, vattr.va_mode); 2668 if (error == 0) 2669 #endif 2670 error = VOP_SETATTR(vp, &vattr, cred); 2671 VOP_UNLOCK(vp, 0); 2672 vn_finished_write(mp); 2673 return (error); 2674 } 2675 2676 /* 2677 * Change mode of a file given path name. 2678 */ 2679 #ifndef _SYS_SYSPROTO_H_ 2680 struct chmod_args { 2681 char *path; 2682 int mode; 2683 }; 2684 #endif 2685 int 2686 sys_chmod(struct thread *td, struct chmod_args *uap) 2687 { 2688 2689 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2690 uap->mode, 0)); 2691 } 2692 2693 #ifndef _SYS_SYSPROTO_H_ 2694 struct fchmodat_args { 2695 int dirfd; 2696 char *path; 2697 mode_t mode; 2698 int flag; 2699 } 2700 #endif 2701 int 2702 sys_fchmodat(struct thread *td, struct fchmodat_args *uap) 2703 { 2704 int flag = uap->flag; 2705 int fd = uap->fd; 2706 char *path = uap->path; 2707 mode_t mode = uap->mode; 2708 2709 if (flag & ~AT_SYMLINK_NOFOLLOW) 2710 return (EINVAL); 2711 2712 return (kern_fchmodat(td, fd, path, UIO_USERSPACE, mode, flag)); 2713 } 2714 2715 /* 2716 * Change mode of a file given path name (don't follow links.) 2717 */ 2718 #ifndef _SYS_SYSPROTO_H_ 2719 struct lchmod_args { 2720 char *path; 2721 int mode; 2722 }; 2723 #endif 2724 int 2725 sys_lchmod(struct thread *td, struct lchmod_args *uap) 2726 { 2727 2728 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2729 uap->mode, AT_SYMLINK_NOFOLLOW)); 2730 } 2731 2732 int 2733 kern_fchmodat(struct thread *td, int fd, char *path, enum uio_seg pathseg, 2734 mode_t mode, int flag) 2735 { 2736 struct nameidata nd; 2737 cap_rights_t rights; 2738 int error, follow; 2739 2740 AUDIT_ARG_MODE(mode); 2741 follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW; 2742 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd, 2743 cap_rights_init(&rights, CAP_FCHMOD), td); 2744 if ((error = namei(&nd)) != 0) 2745 return (error); 2746 NDFREE(&nd, NDF_ONLY_PNBUF); 2747 error = setfmode(td, td->td_ucred, nd.ni_vp, mode); 2748 vrele(nd.ni_vp); 2749 return (error); 2750 } 2751 2752 /* 2753 * Change mode of a file given a file descriptor. 2754 */ 2755 #ifndef _SYS_SYSPROTO_H_ 2756 struct fchmod_args { 2757 int fd; 2758 int mode; 2759 }; 2760 #endif 2761 int 2762 sys_fchmod(struct thread *td, struct fchmod_args *uap) 2763 { 2764 struct file *fp; 2765 cap_rights_t rights; 2766 int error; 2767 2768 AUDIT_ARG_FD(uap->fd); 2769 AUDIT_ARG_MODE(uap->mode); 2770 2771 error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHMOD), &fp); 2772 if (error != 0) 2773 return (error); 2774 error = fo_chmod(fp, uap->mode, td->td_ucred, td); 2775 fdrop(fp, td); 2776 return (error); 2777 } 2778 2779 /* 2780 * Common implementation for chown(), lchown(), and fchown() 2781 */ 2782 int 2783 setfown(struct thread *td, struct ucred *cred, struct vnode *vp, uid_t uid, 2784 gid_t gid) 2785 { 2786 struct mount *mp; 2787 struct vattr vattr; 2788 int error; 2789 2790 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) 2791 return (error); 2792 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2793 VATTR_NULL(&vattr); 2794 vattr.va_uid = uid; 2795 vattr.va_gid = gid; 2796 #ifdef MAC 2797 error = mac_vnode_check_setowner(cred, vp, vattr.va_uid, 2798 vattr.va_gid); 2799 if (error == 0) 2800 #endif 2801 error = VOP_SETATTR(vp, &vattr, cred); 2802 VOP_UNLOCK(vp, 0); 2803 vn_finished_write(mp); 2804 return (error); 2805 } 2806 2807 /* 2808 * Set ownership given a path name. 2809 */ 2810 #ifndef _SYS_SYSPROTO_H_ 2811 struct chown_args { 2812 char *path; 2813 int uid; 2814 int gid; 2815 }; 2816 #endif 2817 int 2818 sys_chown(struct thread *td, struct chown_args *uap) 2819 { 2820 2821 return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid, 2822 uap->gid, 0)); 2823 } 2824 2825 #ifndef _SYS_SYSPROTO_H_ 2826 struct fchownat_args { 2827 int fd; 2828 const char * path; 2829 uid_t uid; 2830 gid_t gid; 2831 int flag; 2832 }; 2833 #endif 2834 int 2835 sys_fchownat(struct thread *td, struct fchownat_args *uap) 2836 { 2837 int flag; 2838 2839 flag = uap->flag; 2840 if (flag & ~AT_SYMLINK_NOFOLLOW) 2841 return (EINVAL); 2842 2843 return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid, 2844 uap->gid, uap->flag)); 2845 } 2846 2847 int 2848 kern_fchownat(struct thread *td, int fd, char *path, enum uio_seg pathseg, 2849 int uid, int gid, int flag) 2850 { 2851 struct nameidata nd; 2852 cap_rights_t rights; 2853 int error, follow; 2854 2855 AUDIT_ARG_OWNER(uid, gid); 2856 follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW; 2857 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd, 2858 cap_rights_init(&rights, CAP_FCHOWN), td); 2859 2860 if ((error = namei(&nd)) != 0) 2861 return (error); 2862 NDFREE(&nd, NDF_ONLY_PNBUF); 2863 error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid); 2864 vrele(nd.ni_vp); 2865 return (error); 2866 } 2867 2868 /* 2869 * Set ownership given a path name, do not cross symlinks. 2870 */ 2871 #ifndef _SYS_SYSPROTO_H_ 2872 struct lchown_args { 2873 char *path; 2874 int uid; 2875 int gid; 2876 }; 2877 #endif 2878 int 2879 sys_lchown(struct thread *td, struct lchown_args *uap) 2880 { 2881 2882 return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2883 uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW)); 2884 } 2885 2886 /* 2887 * Set ownership given a file descriptor. 2888 */ 2889 #ifndef _SYS_SYSPROTO_H_ 2890 struct fchown_args { 2891 int fd; 2892 int uid; 2893 int gid; 2894 }; 2895 #endif 2896 int 2897 sys_fchown(struct thread *td, struct fchown_args *uap) 2898 { 2899 struct file *fp; 2900 cap_rights_t rights; 2901 int error; 2902 2903 AUDIT_ARG_FD(uap->fd); 2904 AUDIT_ARG_OWNER(uap->uid, uap->gid); 2905 error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHOWN), &fp); 2906 if (error != 0) 2907 return (error); 2908 error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td); 2909 fdrop(fp, td); 2910 return (error); 2911 } 2912 2913 /* 2914 * Common implementation code for utimes(), lutimes(), and futimes(). 2915 */ 2916 static int 2917 getutimes(const struct timeval *usrtvp, enum uio_seg tvpseg, 2918 struct timespec *tsp) 2919 { 2920 struct timeval tv[2]; 2921 const struct timeval *tvp; 2922 int error; 2923 2924 if (usrtvp == NULL) { 2925 vfs_timestamp(&tsp[0]); 2926 tsp[1] = tsp[0]; 2927 } else { 2928 if (tvpseg == UIO_SYSSPACE) { 2929 tvp = usrtvp; 2930 } else { 2931 if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0) 2932 return (error); 2933 tvp = tv; 2934 } 2935 2936 if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 || 2937 tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000) 2938 return (EINVAL); 2939 TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]); 2940 TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]); 2941 } 2942 return (0); 2943 } 2944 2945 /* 2946 * Common implementation code for futimens(), utimensat(). 2947 */ 2948 #define UTIMENS_NULL 0x1 2949 #define UTIMENS_EXIT 0x2 2950 static int 2951 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg, 2952 struct timespec *tsp, int *retflags) 2953 { 2954 struct timespec tsnow; 2955 int error; 2956 2957 vfs_timestamp(&tsnow); 2958 *retflags = 0; 2959 if (usrtsp == NULL) { 2960 tsp[0] = tsnow; 2961 tsp[1] = tsnow; 2962 *retflags |= UTIMENS_NULL; 2963 return (0); 2964 } 2965 if (tspseg == UIO_SYSSPACE) { 2966 tsp[0] = usrtsp[0]; 2967 tsp[1] = usrtsp[1]; 2968 } else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0) 2969 return (error); 2970 if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT) 2971 *retflags |= UTIMENS_EXIT; 2972 if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW) 2973 *retflags |= UTIMENS_NULL; 2974 if (tsp[0].tv_nsec == UTIME_OMIT) 2975 tsp[0].tv_sec = VNOVAL; 2976 else if (tsp[0].tv_nsec == UTIME_NOW) 2977 tsp[0] = tsnow; 2978 else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L) 2979 return (EINVAL); 2980 if (tsp[1].tv_nsec == UTIME_OMIT) 2981 tsp[1].tv_sec = VNOVAL; 2982 else if (tsp[1].tv_nsec == UTIME_NOW) 2983 tsp[1] = tsnow; 2984 else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L) 2985 return (EINVAL); 2986 2987 return (0); 2988 } 2989 2990 /* 2991 * Common implementation code for utimes(), lutimes(), futimes(), futimens(), 2992 * and utimensat(). 2993 */ 2994 static int 2995 setutimes(struct thread *td, struct vnode *vp, const struct timespec *ts, 2996 int numtimes, int nullflag) 2997 { 2998 struct mount *mp; 2999 struct vattr vattr; 3000 int error, setbirthtime; 3001 3002 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) 3003 return (error); 3004 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3005 setbirthtime = 0; 3006 if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) && 3007 timespeccmp(&ts[1], &vattr.va_birthtime, < )) 3008 setbirthtime = 1; 3009 VATTR_NULL(&vattr); 3010 vattr.va_atime = ts[0]; 3011 vattr.va_mtime = ts[1]; 3012 if (setbirthtime) 3013 vattr.va_birthtime = ts[1]; 3014 if (numtimes > 2) 3015 vattr.va_birthtime = ts[2]; 3016 if (nullflag) 3017 vattr.va_vaflags |= VA_UTIMES_NULL; 3018 #ifdef MAC 3019 error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime, 3020 vattr.va_mtime); 3021 #endif 3022 if (error == 0) 3023 error = VOP_SETATTR(vp, &vattr, td->td_ucred); 3024 VOP_UNLOCK(vp, 0); 3025 vn_finished_write(mp); 3026 return (error); 3027 } 3028 3029 /* 3030 * Set the access and modification times of a file. 3031 */ 3032 #ifndef _SYS_SYSPROTO_H_ 3033 struct utimes_args { 3034 char *path; 3035 struct timeval *tptr; 3036 }; 3037 #endif 3038 int 3039 sys_utimes(struct thread *td, struct utimes_args *uap) 3040 { 3041 3042 return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 3043 uap->tptr, UIO_USERSPACE)); 3044 } 3045 3046 #ifndef _SYS_SYSPROTO_H_ 3047 struct futimesat_args { 3048 int fd; 3049 const char * path; 3050 const struct timeval * times; 3051 }; 3052 #endif 3053 int 3054 sys_futimesat(struct thread *td, struct futimesat_args *uap) 3055 { 3056 3057 return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE, 3058 uap->times, UIO_USERSPACE)); 3059 } 3060 3061 int 3062 kern_utimesat(struct thread *td, int fd, char *path, enum uio_seg pathseg, 3063 struct timeval *tptr, enum uio_seg tptrseg) 3064 { 3065 struct nameidata nd; 3066 struct timespec ts[2]; 3067 cap_rights_t rights; 3068 int error; 3069 3070 if ((error = getutimes(tptr, tptrseg, ts)) != 0) 3071 return (error); 3072 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd, 3073 cap_rights_init(&rights, CAP_FUTIMES), td); 3074 3075 if ((error = namei(&nd)) != 0) 3076 return (error); 3077 NDFREE(&nd, NDF_ONLY_PNBUF); 3078 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL); 3079 vrele(nd.ni_vp); 3080 return (error); 3081 } 3082 3083 /* 3084 * Set the access and modification times of a file. 3085 */ 3086 #ifndef _SYS_SYSPROTO_H_ 3087 struct lutimes_args { 3088 char *path; 3089 struct timeval *tptr; 3090 }; 3091 #endif 3092 int 3093 sys_lutimes(struct thread *td, struct lutimes_args *uap) 3094 { 3095 3096 return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr, 3097 UIO_USERSPACE)); 3098 } 3099 3100 int 3101 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg, 3102 struct timeval *tptr, enum uio_seg tptrseg) 3103 { 3104 struct timespec ts[2]; 3105 struct nameidata nd; 3106 int error; 3107 3108 if ((error = getutimes(tptr, tptrseg, ts)) != 0) 3109 return (error); 3110 NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td); 3111 if ((error = namei(&nd)) != 0) 3112 return (error); 3113 NDFREE(&nd, NDF_ONLY_PNBUF); 3114 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL); 3115 vrele(nd.ni_vp); 3116 return (error); 3117 } 3118 3119 /* 3120 * Set the access and modification times of a file. 3121 */ 3122 #ifndef _SYS_SYSPROTO_H_ 3123 struct futimes_args { 3124 int fd; 3125 struct timeval *tptr; 3126 }; 3127 #endif 3128 int 3129 sys_futimes(struct thread *td, struct futimes_args *uap) 3130 { 3131 3132 return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE)); 3133 } 3134 3135 int 3136 kern_futimes(struct thread *td, int fd, struct timeval *tptr, 3137 enum uio_seg tptrseg) 3138 { 3139 struct timespec ts[2]; 3140 struct file *fp; 3141 cap_rights_t rights; 3142 int error; 3143 3144 AUDIT_ARG_FD(fd); 3145 error = getutimes(tptr, tptrseg, ts); 3146 if (error != 0) 3147 return (error); 3148 error = getvnode(td, fd, cap_rights_init(&rights, CAP_FUTIMES), &fp); 3149 if (error != 0) 3150 return (error); 3151 #ifdef AUDIT 3152 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY); 3153 AUDIT_ARG_VNODE1(fp->f_vnode); 3154 VOP_UNLOCK(fp->f_vnode, 0); 3155 #endif 3156 error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL); 3157 fdrop(fp, td); 3158 return (error); 3159 } 3160 3161 int 3162 sys_futimens(struct thread *td, struct futimens_args *uap) 3163 { 3164 3165 return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE)); 3166 } 3167 3168 int 3169 kern_futimens(struct thread *td, int fd, struct timespec *tptr, 3170 enum uio_seg tptrseg) 3171 { 3172 struct timespec ts[2]; 3173 struct file *fp; 3174 cap_rights_t rights; 3175 int error, flags; 3176 3177 AUDIT_ARG_FD(fd); 3178 error = getutimens(tptr, tptrseg, ts, &flags); 3179 if (error != 0) 3180 return (error); 3181 if (flags & UTIMENS_EXIT) 3182 return (0); 3183 error = getvnode(td, fd, cap_rights_init(&rights, CAP_FUTIMES), &fp); 3184 if (error != 0) 3185 return (error); 3186 #ifdef AUDIT 3187 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY); 3188 AUDIT_ARG_VNODE1(fp->f_vnode); 3189 VOP_UNLOCK(fp->f_vnode, 0); 3190 #endif 3191 error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL); 3192 fdrop(fp, td); 3193 return (error); 3194 } 3195 3196 int 3197 sys_utimensat(struct thread *td, struct utimensat_args *uap) 3198 { 3199 3200 return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE, 3201 uap->times, UIO_USERSPACE, uap->flag)); 3202 } 3203 3204 int 3205 kern_utimensat(struct thread *td, int fd, char *path, enum uio_seg pathseg, 3206 struct timespec *tptr, enum uio_seg tptrseg, int flag) 3207 { 3208 struct nameidata nd; 3209 struct timespec ts[2]; 3210 cap_rights_t rights; 3211 int error, flags; 3212 3213 if (flag & ~AT_SYMLINK_NOFOLLOW) 3214 return (EINVAL); 3215 3216 if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0) 3217 return (error); 3218 NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : 3219 FOLLOW) | AUDITVNODE1, pathseg, path, fd, 3220 cap_rights_init(&rights, CAP_FUTIMES), td); 3221 if ((error = namei(&nd)) != 0) 3222 return (error); 3223 /* 3224 * We are allowed to call namei() regardless of 2xUTIME_OMIT. 3225 * POSIX states: 3226 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected." 3227 * "Search permission is denied by a component of the path prefix." 3228 */ 3229 NDFREE(&nd, NDF_ONLY_PNBUF); 3230 if ((flags & UTIMENS_EXIT) == 0) 3231 error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL); 3232 vrele(nd.ni_vp); 3233 return (error); 3234 } 3235 3236 /* 3237 * Truncate a file given its path name. 3238 */ 3239 #ifndef _SYS_SYSPROTO_H_ 3240 struct truncate_args { 3241 char *path; 3242 int pad; 3243 off_t length; 3244 }; 3245 #endif 3246 int 3247 sys_truncate(struct thread *td, struct truncate_args *uap) 3248 { 3249 3250 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length)); 3251 } 3252 3253 int 3254 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length) 3255 { 3256 struct mount *mp; 3257 struct vnode *vp; 3258 void *rl_cookie; 3259 struct vattr vattr; 3260 struct nameidata nd; 3261 int error; 3262 3263 if (length < 0) 3264 return(EINVAL); 3265 NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td); 3266 if ((error = namei(&nd)) != 0) 3267 return (error); 3268 vp = nd.ni_vp; 3269 rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX); 3270 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) { 3271 vn_rangelock_unlock(vp, rl_cookie); 3272 vrele(vp); 3273 return (error); 3274 } 3275 NDFREE(&nd, NDF_ONLY_PNBUF); 3276 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3277 if (vp->v_type == VDIR) 3278 error = EISDIR; 3279 #ifdef MAC 3280 else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) { 3281 } 3282 #endif 3283 else if ((error = vn_writechk(vp)) == 0 && 3284 (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) { 3285 VATTR_NULL(&vattr); 3286 vattr.va_size = length; 3287 error = VOP_SETATTR(vp, &vattr, td->td_ucred); 3288 } 3289 VOP_UNLOCK(vp, 0); 3290 vn_finished_write(mp); 3291 vn_rangelock_unlock(vp, rl_cookie); 3292 vrele(vp); 3293 return (error); 3294 } 3295 3296 #if defined(COMPAT_43) 3297 /* 3298 * Truncate a file given its path name. 3299 */ 3300 #ifndef _SYS_SYSPROTO_H_ 3301 struct otruncate_args { 3302 char *path; 3303 long length; 3304 }; 3305 #endif 3306 int 3307 otruncate(struct thread *td, struct otruncate_args *uap) 3308 { 3309 3310 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length)); 3311 } 3312 #endif /* COMPAT_43 */ 3313 3314 #if defined(COMPAT_FREEBSD6) 3315 /* Versions with the pad argument */ 3316 int 3317 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap) 3318 { 3319 3320 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length)); 3321 } 3322 3323 int 3324 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap) 3325 { 3326 3327 return (kern_ftruncate(td, uap->fd, uap->length)); 3328 } 3329 #endif 3330 3331 int 3332 kern_fsync(struct thread *td, int fd, bool fullsync) 3333 { 3334 struct vnode *vp; 3335 struct mount *mp; 3336 struct file *fp; 3337 cap_rights_t rights; 3338 int error, lock_flags; 3339 3340 AUDIT_ARG_FD(fd); 3341 error = getvnode(td, fd, cap_rights_init(&rights, CAP_FSYNC), &fp); 3342 if (error != 0) 3343 return (error); 3344 vp = fp->f_vnode; 3345 #if 0 3346 if (!fullsync) 3347 /* XXXKIB: compete outstanding aio writes */; 3348 #endif 3349 error = vn_start_write(vp, &mp, V_WAIT | PCATCH); 3350 if (error != 0) 3351 goto drop; 3352 if (MNT_SHARED_WRITES(mp) || 3353 ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) { 3354 lock_flags = LK_SHARED; 3355 } else { 3356 lock_flags = LK_EXCLUSIVE; 3357 } 3358 vn_lock(vp, lock_flags | LK_RETRY); 3359 AUDIT_ARG_VNODE1(vp); 3360 if (vp->v_object != NULL) { 3361 VM_OBJECT_WLOCK(vp->v_object); 3362 vm_object_page_clean(vp->v_object, 0, 0, 0); 3363 VM_OBJECT_WUNLOCK(vp->v_object); 3364 } 3365 error = fullsync ? VOP_FSYNC(vp, MNT_WAIT, td) : VOP_FDATASYNC(vp, td); 3366 VOP_UNLOCK(vp, 0); 3367 vn_finished_write(mp); 3368 drop: 3369 fdrop(fp, td); 3370 return (error); 3371 } 3372 3373 /* 3374 * Sync an open file. 3375 */ 3376 #ifndef _SYS_SYSPROTO_H_ 3377 struct fsync_args { 3378 int fd; 3379 }; 3380 #endif 3381 int 3382 sys_fsync(struct thread *td, struct fsync_args *uap) 3383 { 3384 3385 return (kern_fsync(td, uap->fd, true)); 3386 } 3387 3388 int 3389 sys_fdatasync(struct thread *td, struct fdatasync_args *uap) 3390 { 3391 3392 return (kern_fsync(td, uap->fd, false)); 3393 } 3394 3395 /* 3396 * Rename files. Source and destination must either both be directories, or 3397 * both not be directories. If target is a directory, it must be empty. 3398 */ 3399 #ifndef _SYS_SYSPROTO_H_ 3400 struct rename_args { 3401 char *from; 3402 char *to; 3403 }; 3404 #endif 3405 int 3406 sys_rename(struct thread *td, struct rename_args *uap) 3407 { 3408 3409 return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD, 3410 uap->to, UIO_USERSPACE)); 3411 } 3412 3413 #ifndef _SYS_SYSPROTO_H_ 3414 struct renameat_args { 3415 int oldfd; 3416 char *old; 3417 int newfd; 3418 char *new; 3419 }; 3420 #endif 3421 int 3422 sys_renameat(struct thread *td, struct renameat_args *uap) 3423 { 3424 3425 return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new, 3426 UIO_USERSPACE)); 3427 } 3428 3429 int 3430 kern_renameat(struct thread *td, int oldfd, char *old, int newfd, char *new, 3431 enum uio_seg pathseg) 3432 { 3433 struct mount *mp = NULL; 3434 struct vnode *tvp, *fvp, *tdvp; 3435 struct nameidata fromnd, tond; 3436 cap_rights_t rights; 3437 int error; 3438 3439 again: 3440 bwillwrite(); 3441 #ifdef MAC 3442 NDINIT_ATRIGHTS(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART | 3443 AUDITVNODE1, pathseg, old, oldfd, 3444 cap_rights_init(&rights, CAP_RENAMEAT_SOURCE), td); 3445 #else 3446 NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1, 3447 pathseg, old, oldfd, 3448 cap_rights_init(&rights, CAP_RENAMEAT_SOURCE), td); 3449 #endif 3450 3451 if ((error = namei(&fromnd)) != 0) 3452 return (error); 3453 #ifdef MAC 3454 error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp, 3455 fromnd.ni_vp, &fromnd.ni_cnd); 3456 VOP_UNLOCK(fromnd.ni_dvp, 0); 3457 if (fromnd.ni_dvp != fromnd.ni_vp) 3458 VOP_UNLOCK(fromnd.ni_vp, 0); 3459 #endif 3460 fvp = fromnd.ni_vp; 3461 NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | 3462 SAVESTART | AUDITVNODE2, pathseg, new, newfd, 3463 cap_rights_init(&rights, CAP_RENAMEAT_TARGET), td); 3464 if (fromnd.ni_vp->v_type == VDIR) 3465 tond.ni_cnd.cn_flags |= WILLBEDIR; 3466 if ((error = namei(&tond)) != 0) { 3467 /* Translate error code for rename("dir1", "dir2/."). */ 3468 if (error == EISDIR && fvp->v_type == VDIR) 3469 error = EINVAL; 3470 NDFREE(&fromnd, NDF_ONLY_PNBUF); 3471 vrele(fromnd.ni_dvp); 3472 vrele(fvp); 3473 goto out1; 3474 } 3475 tdvp = tond.ni_dvp; 3476 tvp = tond.ni_vp; 3477 error = vn_start_write(fvp, &mp, V_NOWAIT); 3478 if (error != 0) { 3479 NDFREE(&fromnd, NDF_ONLY_PNBUF); 3480 NDFREE(&tond, NDF_ONLY_PNBUF); 3481 if (tvp != NULL) 3482 vput(tvp); 3483 if (tdvp == tvp) 3484 vrele(tdvp); 3485 else 3486 vput(tdvp); 3487 vrele(fromnd.ni_dvp); 3488 vrele(fvp); 3489 vrele(tond.ni_startdir); 3490 if (fromnd.ni_startdir != NULL) 3491 vrele(fromnd.ni_startdir); 3492 error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH); 3493 if (error != 0) 3494 return (error); 3495 goto again; 3496 } 3497 if (tvp != NULL) { 3498 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3499 error = ENOTDIR; 3500 goto out; 3501 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3502 error = EISDIR; 3503 goto out; 3504 } 3505 #ifdef CAPABILITIES 3506 if (newfd != AT_FDCWD) { 3507 /* 3508 * If the target already exists we require CAP_UNLINKAT 3509 * from 'newfd'. 3510 */ 3511 error = cap_check(&tond.ni_filecaps.fc_rights, 3512 cap_rights_init(&rights, CAP_UNLINKAT)); 3513 if (error != 0) 3514 goto out; 3515 } 3516 #endif 3517 } 3518 if (fvp == tdvp) { 3519 error = EINVAL; 3520 goto out; 3521 } 3522 /* 3523 * If the source is the same as the destination (that is, if they 3524 * are links to the same vnode), then there is nothing to do. 3525 */ 3526 if (fvp == tvp) 3527 error = -1; 3528 #ifdef MAC 3529 else 3530 error = mac_vnode_check_rename_to(td->td_ucred, tdvp, 3531 tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd); 3532 #endif 3533 out: 3534 if (error == 0) { 3535 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3536 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3537 NDFREE(&fromnd, NDF_ONLY_PNBUF); 3538 NDFREE(&tond, NDF_ONLY_PNBUF); 3539 } else { 3540 NDFREE(&fromnd, NDF_ONLY_PNBUF); 3541 NDFREE(&tond, NDF_ONLY_PNBUF); 3542 if (tvp != NULL) 3543 vput(tvp); 3544 if (tdvp == tvp) 3545 vrele(tdvp); 3546 else 3547 vput(tdvp); 3548 vrele(fromnd.ni_dvp); 3549 vrele(fvp); 3550 } 3551 vrele(tond.ni_startdir); 3552 vn_finished_write(mp); 3553 out1: 3554 if (fromnd.ni_startdir) 3555 vrele(fromnd.ni_startdir); 3556 if (error == -1) 3557 return (0); 3558 return (error); 3559 } 3560 3561 /* 3562 * Make a directory file. 3563 */ 3564 #ifndef _SYS_SYSPROTO_H_ 3565 struct mkdir_args { 3566 char *path; 3567 int mode; 3568 }; 3569 #endif 3570 int 3571 sys_mkdir(struct thread *td, struct mkdir_args *uap) 3572 { 3573 3574 return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 3575 uap->mode)); 3576 } 3577 3578 #ifndef _SYS_SYSPROTO_H_ 3579 struct mkdirat_args { 3580 int fd; 3581 char *path; 3582 mode_t mode; 3583 }; 3584 #endif 3585 int 3586 sys_mkdirat(struct thread *td, struct mkdirat_args *uap) 3587 { 3588 3589 return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode)); 3590 } 3591 3592 int 3593 kern_mkdirat(struct thread *td, int fd, char *path, enum uio_seg segflg, 3594 int mode) 3595 { 3596 struct mount *mp; 3597 struct vnode *vp; 3598 struct vattr vattr; 3599 struct nameidata nd; 3600 cap_rights_t rights; 3601 int error; 3602 3603 AUDIT_ARG_MODE(mode); 3604 restart: 3605 bwillwrite(); 3606 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 | 3607 NOCACHE, segflg, path, fd, cap_rights_init(&rights, CAP_MKDIRAT), 3608 td); 3609 nd.ni_cnd.cn_flags |= WILLBEDIR; 3610 if ((error = namei(&nd)) != 0) 3611 return (error); 3612 vp = nd.ni_vp; 3613 if (vp != NULL) { 3614 NDFREE(&nd, NDF_ONLY_PNBUF); 3615 /* 3616 * XXX namei called with LOCKPARENT but not LOCKLEAF has 3617 * the strange behaviour of leaving the vnode unlocked 3618 * if the target is the same vnode as the parent. 3619 */ 3620 if (vp == nd.ni_dvp) 3621 vrele(nd.ni_dvp); 3622 else 3623 vput(nd.ni_dvp); 3624 vrele(vp); 3625 return (EEXIST); 3626 } 3627 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3628 NDFREE(&nd, NDF_ONLY_PNBUF); 3629 vput(nd.ni_dvp); 3630 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0) 3631 return (error); 3632 goto restart; 3633 } 3634 VATTR_NULL(&vattr); 3635 vattr.va_type = VDIR; 3636 vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask; 3637 #ifdef MAC 3638 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd, 3639 &vattr); 3640 if (error != 0) 3641 goto out; 3642 #endif 3643 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3644 #ifdef MAC 3645 out: 3646 #endif 3647 NDFREE(&nd, NDF_ONLY_PNBUF); 3648 vput(nd.ni_dvp); 3649 if (error == 0) 3650 vput(nd.ni_vp); 3651 vn_finished_write(mp); 3652 return (error); 3653 } 3654 3655 /* 3656 * Remove a directory file. 3657 */ 3658 #ifndef _SYS_SYSPROTO_H_ 3659 struct rmdir_args { 3660 char *path; 3661 }; 3662 #endif 3663 int 3664 sys_rmdir(struct thread *td, struct rmdir_args *uap) 3665 { 3666 3667 return (kern_rmdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE)); 3668 } 3669 3670 int 3671 kern_rmdirat(struct thread *td, int fd, char *path, enum uio_seg pathseg) 3672 { 3673 struct mount *mp; 3674 struct vnode *vp; 3675 struct nameidata nd; 3676 cap_rights_t rights; 3677 int error; 3678 3679 restart: 3680 bwillwrite(); 3681 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1, 3682 pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td); 3683 if ((error = namei(&nd)) != 0) 3684 return (error); 3685 vp = nd.ni_vp; 3686 if (vp->v_type != VDIR) { 3687 error = ENOTDIR; 3688 goto out; 3689 } 3690 /* 3691 * No rmdir "." please. 3692 */ 3693 if (nd.ni_dvp == vp) { 3694 error = EINVAL; 3695 goto out; 3696 } 3697 /* 3698 * The root of a mounted filesystem cannot be deleted. 3699 */ 3700 if (vp->v_vflag & VV_ROOT) { 3701 error = EBUSY; 3702 goto out; 3703 } 3704 #ifdef MAC 3705 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp, 3706 &nd.ni_cnd); 3707 if (error != 0) 3708 goto out; 3709 #endif 3710 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3711 NDFREE(&nd, NDF_ONLY_PNBUF); 3712 vput(vp); 3713 if (nd.ni_dvp == vp) 3714 vrele(nd.ni_dvp); 3715 else 3716 vput(nd.ni_dvp); 3717 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0) 3718 return (error); 3719 goto restart; 3720 } 3721 vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK); 3722 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3723 vn_finished_write(mp); 3724 out: 3725 NDFREE(&nd, NDF_ONLY_PNBUF); 3726 vput(vp); 3727 if (nd.ni_dvp == vp) 3728 vrele(nd.ni_dvp); 3729 else 3730 vput(nd.ni_dvp); 3731 return (error); 3732 } 3733 3734 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11) 3735 int 3736 freebsd11_kern_getdirentries(struct thread *td, int fd, char *ubuf, u_int count, 3737 long *basep, void (*func)(struct freebsd11_dirent *)) 3738 { 3739 struct freebsd11_dirent dstdp; 3740 struct dirent *dp, *edp; 3741 char *dirbuf; 3742 off_t base; 3743 ssize_t resid, ucount; 3744 int error; 3745 3746 /* XXX arbitrary sanity limit on `count'. */ 3747 count = min(count, 64 * 1024); 3748 3749 dirbuf = malloc(count, M_TEMP, M_WAITOK); 3750 3751 error = kern_getdirentries(td, fd, dirbuf, count, &base, &resid, 3752 UIO_SYSSPACE); 3753 if (error != 0) 3754 goto done; 3755 if (basep != NULL) 3756 *basep = base; 3757 3758 ucount = 0; 3759 for (dp = (struct dirent *)dirbuf, 3760 edp = (struct dirent *)&dirbuf[count - resid]; 3761 ucount < count && dp < edp; ) { 3762 if (dp->d_reclen == 0) 3763 break; 3764 MPASS(dp->d_reclen >= _GENERIC_DIRLEN(0)); 3765 if (dp->d_namlen >= sizeof(dstdp.d_name)) 3766 continue; 3767 dstdp.d_type = dp->d_type; 3768 dstdp.d_namlen = dp->d_namlen; 3769 dstdp.d_fileno = dp->d_fileno; /* truncate */ 3770 if (dstdp.d_fileno != dp->d_fileno) { 3771 switch (ino64_trunc_error) { 3772 default: 3773 case 0: 3774 break; 3775 case 1: 3776 error = EOVERFLOW; 3777 goto done; 3778 case 2: 3779 dstdp.d_fileno = UINT32_MAX; 3780 break; 3781 } 3782 } 3783 dstdp.d_reclen = sizeof(dstdp) - sizeof(dstdp.d_name) + 3784 ((dp->d_namlen + 1 + 3) &~ 3); 3785 bcopy(dp->d_name, dstdp.d_name, dstdp.d_namlen); 3786 bzero(dstdp.d_name + dstdp.d_namlen, 3787 dstdp.d_reclen - offsetof(struct freebsd11_dirent, d_name) - 3788 dstdp.d_namlen); 3789 MPASS(dstdp.d_reclen <= dp->d_reclen); 3790 MPASS(ucount + dstdp.d_reclen <= count); 3791 if (func != NULL) 3792 func(&dstdp); 3793 error = copyout(&dstdp, ubuf + ucount, dstdp.d_reclen); 3794 if (error != 0) 3795 break; 3796 dp = (struct dirent *)((char *)dp + dp->d_reclen); 3797 ucount += dstdp.d_reclen; 3798 } 3799 3800 done: 3801 free(dirbuf, M_TEMP); 3802 if (error == 0) 3803 td->td_retval[0] = ucount; 3804 return (error); 3805 } 3806 #endif /* COMPAT */ 3807 3808 #ifdef COMPAT_43 3809 static void 3810 ogetdirentries_cvt(struct freebsd11_dirent *dp) 3811 { 3812 #if (BYTE_ORDER == LITTLE_ENDIAN) 3813 /* 3814 * The expected low byte of dp->d_namlen is our dp->d_type. 3815 * The high MBZ byte of dp->d_namlen is our dp->d_namlen. 3816 */ 3817 dp->d_type = dp->d_namlen; 3818 dp->d_namlen = 0; 3819 #else 3820 /* 3821 * The dp->d_type is the high byte of the expected dp->d_namlen, 3822 * so must be zero'ed. 3823 */ 3824 dp->d_type = 0; 3825 #endif 3826 } 3827 3828 /* 3829 * Read a block of directory entries in a filesystem independent format. 3830 */ 3831 #ifndef _SYS_SYSPROTO_H_ 3832 struct ogetdirentries_args { 3833 int fd; 3834 char *buf; 3835 u_int count; 3836 long *basep; 3837 }; 3838 #endif 3839 int 3840 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap) 3841 { 3842 long loff; 3843 int error; 3844 3845 error = kern_ogetdirentries(td, uap, &loff); 3846 if (error == 0) 3847 error = copyout(&loff, uap->basep, sizeof(long)); 3848 return (error); 3849 } 3850 3851 int 3852 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap, 3853 long *ploff) 3854 { 3855 long base; 3856 int error; 3857 3858 /* XXX arbitrary sanity limit on `count'. */ 3859 if (uap->count > 64 * 1024) 3860 return (EINVAL); 3861 3862 error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count, 3863 &base, ogetdirentries_cvt); 3864 3865 if (error == 0 && uap->basep != NULL) 3866 error = copyout(&base, uap->basep, sizeof(long)); 3867 3868 return (error); 3869 } 3870 #endif /* COMPAT_43 */ 3871 3872 #if defined(COMPAT_FREEBSD11) 3873 #ifndef _SYS_SYSPROTO_H_ 3874 struct freebsd11_getdirentries_args { 3875 int fd; 3876 char *buf; 3877 u_int count; 3878 long *basep; 3879 }; 3880 #endif 3881 int 3882 freebsd11_getdirentries(struct thread *td, 3883 struct freebsd11_getdirentries_args *uap) 3884 { 3885 long base; 3886 int error; 3887 3888 error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count, 3889 &base, NULL); 3890 3891 if (error == 0 && uap->basep != NULL) 3892 error = copyout(&base, uap->basep, sizeof(long)); 3893 return (error); 3894 } 3895 3896 int 3897 freebsd11_getdents(struct thread *td, struct freebsd11_getdents_args *uap) 3898 { 3899 struct freebsd11_getdirentries_args ap; 3900 3901 ap.fd = uap->fd; 3902 ap.buf = uap->buf; 3903 ap.count = uap->count; 3904 ap.basep = NULL; 3905 return (freebsd11_getdirentries(td, &ap)); 3906 } 3907 #endif /* COMPAT_FREEBSD11 */ 3908 3909 /* 3910 * Read a block of directory entries in a filesystem independent format. 3911 */ 3912 int 3913 sys_getdirentries(struct thread *td, struct getdirentries_args *uap) 3914 { 3915 off_t base; 3916 int error; 3917 3918 error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base, 3919 NULL, UIO_USERSPACE); 3920 if (error != 0) 3921 return (error); 3922 if (uap->basep != NULL) 3923 error = copyout(&base, uap->basep, sizeof(off_t)); 3924 return (error); 3925 } 3926 3927 int 3928 kern_getdirentries(struct thread *td, int fd, char *buf, size_t count, 3929 off_t *basep, ssize_t *residp, enum uio_seg bufseg) 3930 { 3931 struct vnode *vp; 3932 struct file *fp; 3933 struct uio auio; 3934 struct iovec aiov; 3935 cap_rights_t rights; 3936 off_t loff; 3937 int error, eofflag; 3938 off_t foffset; 3939 3940 AUDIT_ARG_FD(fd); 3941 if (count > IOSIZE_MAX) 3942 return (EINVAL); 3943 auio.uio_resid = count; 3944 error = getvnode(td, fd, cap_rights_init(&rights, CAP_READ), &fp); 3945 if (error != 0) 3946 return (error); 3947 if ((fp->f_flag & FREAD) == 0) { 3948 fdrop(fp, td); 3949 return (EBADF); 3950 } 3951 vp = fp->f_vnode; 3952 foffset = foffset_lock(fp, 0); 3953 unionread: 3954 if (vp->v_type != VDIR) { 3955 error = EINVAL; 3956 goto fail; 3957 } 3958 aiov.iov_base = buf; 3959 aiov.iov_len = count; 3960 auio.uio_iov = &aiov; 3961 auio.uio_iovcnt = 1; 3962 auio.uio_rw = UIO_READ; 3963 auio.uio_segflg = bufseg; 3964 auio.uio_td = td; 3965 vn_lock(vp, LK_SHARED | LK_RETRY); 3966 AUDIT_ARG_VNODE1(vp); 3967 loff = auio.uio_offset = foffset; 3968 #ifdef MAC 3969 error = mac_vnode_check_readdir(td->td_ucred, vp); 3970 if (error == 0) 3971 #endif 3972 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL, 3973 NULL); 3974 foffset = auio.uio_offset; 3975 if (error != 0) { 3976 VOP_UNLOCK(vp, 0); 3977 goto fail; 3978 } 3979 if (count == auio.uio_resid && 3980 (vp->v_vflag & VV_ROOT) && 3981 (vp->v_mount->mnt_flag & MNT_UNION)) { 3982 struct vnode *tvp = vp; 3983 3984 vp = vp->v_mount->mnt_vnodecovered; 3985 VREF(vp); 3986 fp->f_vnode = vp; 3987 fp->f_data = vp; 3988 foffset = 0; 3989 vput(tvp); 3990 goto unionread; 3991 } 3992 VOP_UNLOCK(vp, 0); 3993 *basep = loff; 3994 if (residp != NULL) 3995 *residp = auio.uio_resid; 3996 td->td_retval[0] = count - auio.uio_resid; 3997 fail: 3998 foffset_unlock(fp, foffset, 0); 3999 fdrop(fp, td); 4000 return (error); 4001 } 4002 4003 /* 4004 * Set the mode mask for creation of filesystem nodes. 4005 */ 4006 #ifndef _SYS_SYSPROTO_H_ 4007 struct umask_args { 4008 int newmask; 4009 }; 4010 #endif 4011 int 4012 sys_umask(struct thread *td, struct umask_args *uap) 4013 { 4014 struct filedesc *fdp; 4015 4016 fdp = td->td_proc->p_fd; 4017 FILEDESC_XLOCK(fdp); 4018 td->td_retval[0] = fdp->fd_cmask; 4019 fdp->fd_cmask = uap->newmask & ALLPERMS; 4020 FILEDESC_XUNLOCK(fdp); 4021 return (0); 4022 } 4023 4024 /* 4025 * Void all references to file by ripping underlying filesystem away from 4026 * vnode. 4027 */ 4028 #ifndef _SYS_SYSPROTO_H_ 4029 struct revoke_args { 4030 char *path; 4031 }; 4032 #endif 4033 int 4034 sys_revoke(struct thread *td, struct revoke_args *uap) 4035 { 4036 struct vnode *vp; 4037 struct vattr vattr; 4038 struct nameidata nd; 4039 int error; 4040 4041 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE, 4042 uap->path, td); 4043 if ((error = namei(&nd)) != 0) 4044 return (error); 4045 vp = nd.ni_vp; 4046 NDFREE(&nd, NDF_ONLY_PNBUF); 4047 if (vp->v_type != VCHR || vp->v_rdev == NULL) { 4048 error = EINVAL; 4049 goto out; 4050 } 4051 #ifdef MAC 4052 error = mac_vnode_check_revoke(td->td_ucred, vp); 4053 if (error != 0) 4054 goto out; 4055 #endif 4056 error = VOP_GETATTR(vp, &vattr, td->td_ucred); 4057 if (error != 0) 4058 goto out; 4059 if (td->td_ucred->cr_uid != vattr.va_uid) { 4060 error = priv_check(td, PRIV_VFS_ADMIN); 4061 if (error != 0) 4062 goto out; 4063 } 4064 if (vcount(vp) > 1) 4065 VOP_REVOKE(vp, REVOKEALL); 4066 out: 4067 vput(vp); 4068 return (error); 4069 } 4070 4071 /* 4072 * Convert a user file descriptor to a kernel file entry and check that, if it 4073 * is a capability, the correct rights are present. A reference on the file 4074 * entry is held upon returning. 4075 */ 4076 int 4077 getvnode(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp) 4078 { 4079 struct file *fp; 4080 int error; 4081 4082 error = fget_unlocked(td->td_proc->p_fd, fd, rightsp, &fp, NULL); 4083 if (error != 0) 4084 return (error); 4085 4086 /* 4087 * The file could be not of the vnode type, or it may be not 4088 * yet fully initialized, in which case the f_vnode pointer 4089 * may be set, but f_ops is still badfileops. E.g., 4090 * devfs_open() transiently create such situation to 4091 * facilitate csw d_fdopen(). 4092 * 4093 * Dupfdopen() handling in kern_openat() installs the 4094 * half-baked file into the process descriptor table, allowing 4095 * other thread to dereference it. Guard against the race by 4096 * checking f_ops. 4097 */ 4098 if (fp->f_vnode == NULL || fp->f_ops == &badfileops) { 4099 fdrop(fp, td); 4100 return (EINVAL); 4101 } 4102 *fpp = fp; 4103 return (0); 4104 } 4105 4106 4107 /* 4108 * Get an (NFS) file handle. 4109 */ 4110 #ifndef _SYS_SYSPROTO_H_ 4111 struct lgetfh_args { 4112 char *fname; 4113 fhandle_t *fhp; 4114 }; 4115 #endif 4116 int 4117 sys_lgetfh(struct thread *td, struct lgetfh_args *uap) 4118 { 4119 struct nameidata nd; 4120 fhandle_t fh; 4121 struct vnode *vp; 4122 int error; 4123 4124 error = priv_check(td, PRIV_VFS_GETFH); 4125 if (error != 0) 4126 return (error); 4127 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE, 4128 uap->fname, td); 4129 error = namei(&nd); 4130 if (error != 0) 4131 return (error); 4132 NDFREE(&nd, NDF_ONLY_PNBUF); 4133 vp = nd.ni_vp; 4134 bzero(&fh, sizeof(fh)); 4135 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid; 4136 error = VOP_VPTOFH(vp, &fh.fh_fid); 4137 vput(vp); 4138 if (error == 0) 4139 error = copyout(&fh, uap->fhp, sizeof (fh)); 4140 return (error); 4141 } 4142 4143 #ifndef _SYS_SYSPROTO_H_ 4144 struct getfh_args { 4145 char *fname; 4146 fhandle_t *fhp; 4147 }; 4148 #endif 4149 int 4150 sys_getfh(struct thread *td, struct getfh_args *uap) 4151 { 4152 struct nameidata nd; 4153 fhandle_t fh; 4154 struct vnode *vp; 4155 int error; 4156 4157 error = priv_check(td, PRIV_VFS_GETFH); 4158 if (error != 0) 4159 return (error); 4160 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE, 4161 uap->fname, td); 4162 error = namei(&nd); 4163 if (error != 0) 4164 return (error); 4165 NDFREE(&nd, NDF_ONLY_PNBUF); 4166 vp = nd.ni_vp; 4167 bzero(&fh, sizeof(fh)); 4168 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid; 4169 error = VOP_VPTOFH(vp, &fh.fh_fid); 4170 vput(vp); 4171 if (error == 0) 4172 error = copyout(&fh, uap->fhp, sizeof (fh)); 4173 return (error); 4174 } 4175 4176 /* 4177 * syscall for the rpc.lockd to use to translate a NFS file handle into an 4178 * open descriptor. 4179 * 4180 * warning: do not remove the priv_check() call or this becomes one giant 4181 * security hole. 4182 */ 4183 #ifndef _SYS_SYSPROTO_H_ 4184 struct fhopen_args { 4185 const struct fhandle *u_fhp; 4186 int flags; 4187 }; 4188 #endif 4189 int 4190 sys_fhopen(struct thread *td, struct fhopen_args *uap) 4191 { 4192 struct mount *mp; 4193 struct vnode *vp; 4194 struct fhandle fhp; 4195 struct file *fp; 4196 int fmode, error; 4197 int indx; 4198 4199 error = priv_check(td, PRIV_VFS_FHOPEN); 4200 if (error != 0) 4201 return (error); 4202 indx = -1; 4203 fmode = FFLAGS(uap->flags); 4204 /* why not allow a non-read/write open for our lockd? */ 4205 if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT)) 4206 return (EINVAL); 4207 error = copyin(uap->u_fhp, &fhp, sizeof(fhp)); 4208 if (error != 0) 4209 return(error); 4210 /* find the mount point */ 4211 mp = vfs_busyfs(&fhp.fh_fsid); 4212 if (mp == NULL) 4213 return (ESTALE); 4214 /* now give me my vnode, it gets returned to me locked */ 4215 error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp); 4216 vfs_unbusy(mp); 4217 if (error != 0) 4218 return (error); 4219 4220 error = falloc_noinstall(td, &fp); 4221 if (error != 0) { 4222 vput(vp); 4223 return (error); 4224 } 4225 /* 4226 * An extra reference on `fp' has been held for us by 4227 * falloc_noinstall(). 4228 */ 4229 4230 #ifdef INVARIANTS 4231 td->td_dupfd = -1; 4232 #endif 4233 error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp); 4234 if (error != 0) { 4235 KASSERT(fp->f_ops == &badfileops, 4236 ("VOP_OPEN in fhopen() set f_ops")); 4237 KASSERT(td->td_dupfd < 0, 4238 ("fhopen() encountered fdopen()")); 4239 4240 vput(vp); 4241 goto bad; 4242 } 4243 #ifdef INVARIANTS 4244 td->td_dupfd = 0; 4245 #endif 4246 fp->f_vnode = vp; 4247 fp->f_seqcount = 1; 4248 finit(fp, (fmode & FMASK) | (fp->f_flag & FHASLOCK), DTYPE_VNODE, vp, 4249 &vnops); 4250 VOP_UNLOCK(vp, 0); 4251 if ((fmode & O_TRUNC) != 0) { 4252 error = fo_truncate(fp, 0, td->td_ucred, td); 4253 if (error != 0) 4254 goto bad; 4255 } 4256 4257 error = finstall(td, fp, &indx, fmode, NULL); 4258 bad: 4259 fdrop(fp, td); 4260 td->td_retval[0] = indx; 4261 return (error); 4262 } 4263 4264 /* 4265 * Stat an (NFS) file handle. 4266 */ 4267 #ifndef _SYS_SYSPROTO_H_ 4268 struct fhstat_args { 4269 struct fhandle *u_fhp; 4270 struct stat *sb; 4271 }; 4272 #endif 4273 int 4274 sys_fhstat(struct thread *td, struct fhstat_args *uap) 4275 { 4276 struct stat sb; 4277 struct fhandle fh; 4278 int error; 4279 4280 error = copyin(uap->u_fhp, &fh, sizeof(fh)); 4281 if (error != 0) 4282 return (error); 4283 error = kern_fhstat(td, fh, &sb); 4284 if (error == 0) 4285 error = copyout(&sb, uap->sb, sizeof(sb)); 4286 return (error); 4287 } 4288 4289 int 4290 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb) 4291 { 4292 struct mount *mp; 4293 struct vnode *vp; 4294 int error; 4295 4296 error = priv_check(td, PRIV_VFS_FHSTAT); 4297 if (error != 0) 4298 return (error); 4299 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL) 4300 return (ESTALE); 4301 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp); 4302 vfs_unbusy(mp); 4303 if (error != 0) 4304 return (error); 4305 error = vn_stat(vp, sb, td->td_ucred, NOCRED, td); 4306 vput(vp); 4307 return (error); 4308 } 4309 4310 /* 4311 * Implement fstatfs() for (NFS) file handles. 4312 */ 4313 #ifndef _SYS_SYSPROTO_H_ 4314 struct fhstatfs_args { 4315 struct fhandle *u_fhp; 4316 struct statfs *buf; 4317 }; 4318 #endif 4319 int 4320 sys_fhstatfs(struct thread *td, struct fhstatfs_args *uap) 4321 { 4322 struct statfs *sfp; 4323 fhandle_t fh; 4324 int error; 4325 4326 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t)); 4327 if (error != 0) 4328 return (error); 4329 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 4330 error = kern_fhstatfs(td, fh, sfp); 4331 if (error == 0) 4332 error = copyout(sfp, uap->buf, sizeof(*sfp)); 4333 free(sfp, M_STATFS); 4334 return (error); 4335 } 4336 4337 int 4338 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf) 4339 { 4340 struct statfs *sp; 4341 struct mount *mp; 4342 struct vnode *vp; 4343 int error; 4344 4345 error = priv_check(td, PRIV_VFS_FHSTATFS); 4346 if (error != 0) 4347 return (error); 4348 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL) 4349 return (ESTALE); 4350 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp); 4351 if (error != 0) { 4352 vfs_unbusy(mp); 4353 return (error); 4354 } 4355 vput(vp); 4356 error = prison_canseemount(td->td_ucred, mp); 4357 if (error != 0) 4358 goto out; 4359 #ifdef MAC 4360 error = mac_mount_check_stat(td->td_ucred, mp); 4361 if (error != 0) 4362 goto out; 4363 #endif 4364 /* 4365 * Set these in case the underlying filesystem fails to do so. 4366 */ 4367 sp = &mp->mnt_stat; 4368 sp->f_version = STATFS_VERSION; 4369 sp->f_namemax = NAME_MAX; 4370 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK; 4371 error = VFS_STATFS(mp, sp); 4372 if (error == 0) 4373 *buf = *sp; 4374 out: 4375 vfs_unbusy(mp); 4376 return (error); 4377 } 4378 4379 int 4380 kern_posix_fallocate(struct thread *td, int fd, off_t offset, off_t len) 4381 { 4382 struct file *fp; 4383 struct mount *mp; 4384 struct vnode *vp; 4385 cap_rights_t rights; 4386 off_t olen, ooffset; 4387 int error; 4388 #ifdef AUDIT 4389 int audited_vnode1 = 0; 4390 #endif 4391 4392 AUDIT_ARG_FD(fd); 4393 if (offset < 0 || len <= 0) 4394 return (EINVAL); 4395 /* Check for wrap. */ 4396 if (offset > OFF_MAX - len) 4397 return (EFBIG); 4398 AUDIT_ARG_FD(fd); 4399 error = fget(td, fd, cap_rights_init(&rights, CAP_PWRITE), &fp); 4400 if (error != 0) 4401 return (error); 4402 AUDIT_ARG_FILE(td->td_proc, fp); 4403 if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) { 4404 error = ESPIPE; 4405 goto out; 4406 } 4407 if ((fp->f_flag & FWRITE) == 0) { 4408 error = EBADF; 4409 goto out; 4410 } 4411 if (fp->f_type != DTYPE_VNODE) { 4412 error = ENODEV; 4413 goto out; 4414 } 4415 vp = fp->f_vnode; 4416 if (vp->v_type != VREG) { 4417 error = ENODEV; 4418 goto out; 4419 } 4420 4421 /* Allocating blocks may take a long time, so iterate. */ 4422 for (;;) { 4423 olen = len; 4424 ooffset = offset; 4425 4426 bwillwrite(); 4427 mp = NULL; 4428 error = vn_start_write(vp, &mp, V_WAIT | PCATCH); 4429 if (error != 0) 4430 break; 4431 error = vn_lock(vp, LK_EXCLUSIVE); 4432 if (error != 0) { 4433 vn_finished_write(mp); 4434 break; 4435 } 4436 #ifdef AUDIT 4437 if (!audited_vnode1) { 4438 AUDIT_ARG_VNODE1(vp); 4439 audited_vnode1 = 1; 4440 } 4441 #endif 4442 #ifdef MAC 4443 error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp); 4444 if (error == 0) 4445 #endif 4446 error = VOP_ALLOCATE(vp, &offset, &len); 4447 VOP_UNLOCK(vp, 0); 4448 vn_finished_write(mp); 4449 4450 if (olen + ooffset != offset + len) { 4451 panic("offset + len changed from %jx/%jx to %jx/%jx", 4452 ooffset, olen, offset, len); 4453 } 4454 if (error != 0 || len == 0) 4455 break; 4456 KASSERT(olen > len, ("Iteration did not make progress?")); 4457 maybe_yield(); 4458 } 4459 out: 4460 fdrop(fp, td); 4461 return (error); 4462 } 4463 4464 int 4465 sys_posix_fallocate(struct thread *td, struct posix_fallocate_args *uap) 4466 { 4467 int error; 4468 4469 error = kern_posix_fallocate(td, uap->fd, uap->offset, uap->len); 4470 return (kern_posix_error(td, error)); 4471 } 4472 4473 /* 4474 * Unlike madvise(2), we do not make a best effort to remember every 4475 * possible caching hint. Instead, we remember the last setting with 4476 * the exception that we will allow POSIX_FADV_NORMAL to adjust the 4477 * region of any current setting. 4478 */ 4479 int 4480 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len, 4481 int advice) 4482 { 4483 struct fadvise_info *fa, *new; 4484 struct file *fp; 4485 struct vnode *vp; 4486 cap_rights_t rights; 4487 off_t end; 4488 int error; 4489 4490 if (offset < 0 || len < 0 || offset > OFF_MAX - len) 4491 return (EINVAL); 4492 AUDIT_ARG_VALUE(advice); 4493 switch (advice) { 4494 case POSIX_FADV_SEQUENTIAL: 4495 case POSIX_FADV_RANDOM: 4496 case POSIX_FADV_NOREUSE: 4497 new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK); 4498 break; 4499 case POSIX_FADV_NORMAL: 4500 case POSIX_FADV_WILLNEED: 4501 case POSIX_FADV_DONTNEED: 4502 new = NULL; 4503 break; 4504 default: 4505 return (EINVAL); 4506 } 4507 /* XXX: CAP_POSIX_FADVISE? */ 4508 AUDIT_ARG_FD(fd); 4509 error = fget(td, fd, cap_rights_init(&rights), &fp); 4510 if (error != 0) 4511 goto out; 4512 AUDIT_ARG_FILE(td->td_proc, fp); 4513 if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) { 4514 error = ESPIPE; 4515 goto out; 4516 } 4517 if (fp->f_type != DTYPE_VNODE) { 4518 error = ENODEV; 4519 goto out; 4520 } 4521 vp = fp->f_vnode; 4522 if (vp->v_type != VREG) { 4523 error = ENODEV; 4524 goto out; 4525 } 4526 if (len == 0) 4527 end = OFF_MAX; 4528 else 4529 end = offset + len - 1; 4530 switch (advice) { 4531 case POSIX_FADV_SEQUENTIAL: 4532 case POSIX_FADV_RANDOM: 4533 case POSIX_FADV_NOREUSE: 4534 /* 4535 * Try to merge any existing non-standard region with 4536 * this new region if possible, otherwise create a new 4537 * non-standard region for this request. 4538 */ 4539 mtx_pool_lock(mtxpool_sleep, fp); 4540 fa = fp->f_advice; 4541 if (fa != NULL && fa->fa_advice == advice && 4542 ((fa->fa_start <= end && fa->fa_end >= offset) || 4543 (end != OFF_MAX && fa->fa_start == end + 1) || 4544 (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) { 4545 if (offset < fa->fa_start) 4546 fa->fa_start = offset; 4547 if (end > fa->fa_end) 4548 fa->fa_end = end; 4549 } else { 4550 new->fa_advice = advice; 4551 new->fa_start = offset; 4552 new->fa_end = end; 4553 fp->f_advice = new; 4554 new = fa; 4555 } 4556 mtx_pool_unlock(mtxpool_sleep, fp); 4557 break; 4558 case POSIX_FADV_NORMAL: 4559 /* 4560 * If a the "normal" region overlaps with an existing 4561 * non-standard region, trim or remove the 4562 * non-standard region. 4563 */ 4564 mtx_pool_lock(mtxpool_sleep, fp); 4565 fa = fp->f_advice; 4566 if (fa != NULL) { 4567 if (offset <= fa->fa_start && end >= fa->fa_end) { 4568 new = fa; 4569 fp->f_advice = NULL; 4570 } else if (offset <= fa->fa_start && 4571 end >= fa->fa_start) 4572 fa->fa_start = end + 1; 4573 else if (offset <= fa->fa_end && end >= fa->fa_end) 4574 fa->fa_end = offset - 1; 4575 else if (offset >= fa->fa_start && end <= fa->fa_end) { 4576 /* 4577 * If the "normal" region is a middle 4578 * portion of the existing 4579 * non-standard region, just remove 4580 * the whole thing rather than picking 4581 * one side or the other to 4582 * preserve. 4583 */ 4584 new = fa; 4585 fp->f_advice = NULL; 4586 } 4587 } 4588 mtx_pool_unlock(mtxpool_sleep, fp); 4589 break; 4590 case POSIX_FADV_WILLNEED: 4591 case POSIX_FADV_DONTNEED: 4592 error = VOP_ADVISE(vp, offset, end, advice); 4593 break; 4594 } 4595 out: 4596 if (fp != NULL) 4597 fdrop(fp, td); 4598 free(new, M_FADVISE); 4599 return (error); 4600 } 4601 4602 int 4603 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap) 4604 { 4605 int error; 4606 4607 error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len, 4608 uap->advice); 4609 return (kern_posix_error(td, error)); 4610 } 4611