Annotation of src/sys/kern/vfs_syscalls.c, Revision 1.389
1.389 ! ad 1: /* $NetBSD: vfs_syscalls.c,v 1.388 2009/02/15 03:52:49 enami Exp $ */
1.345 ad 2:
3: /*-
4: * Copyright (c) 2008 The NetBSD Foundation, Inc.
5: * All rights reserved.
6: *
7: * Redistribution and use in source and binary forms, with or without
8: * modification, are permitted provided that the following conditions
9: * are met:
10: * 1. Redistributions of source code must retain the above copyright
11: * notice, this list of conditions and the following disclaimer.
12: * 2. Redistributions in binary form must reproduce the above copyright
13: * notice, this list of conditions and the following disclaimer in the
14: * documentation and/or other materials provided with the distribution.
15: *
16: * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
17: * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
18: * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
19: * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
20: * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
21: * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
22: * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
23: * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
24: * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
25: * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
26: * POSSIBILITY OF SUCH DAMAGE.
27: */
1.31 cgd 28:
29: /*
30: * Copyright (c) 1989, 1993
31: * The Regents of the University of California. All rights reserved.
32: * (c) UNIX System Laboratories, Inc.
33: * All or some portions of this file are derived from material licensed
34: * to the University of California by American Telephone and Telegraph
35: * Co. or Unix System Laboratories, Inc. and are reproduced herein with
36: * the permission of UNIX System Laboratories, Inc.
37: *
38: * Redistribution and use in source and binary forms, with or without
39: * modification, are permitted provided that the following conditions
40: * are met:
41: * 1. Redistributions of source code must retain the above copyright
42: * notice, this list of conditions and the following disclaimer.
43: * 2. Redistributions in binary form must reproduce the above copyright
44: * notice, this list of conditions and the following disclaimer in the
45: * documentation and/or other materials provided with the distribution.
1.191 agc 46: * 3. Neither the name of the University nor the names of its contributors
1.31 cgd 47: * may be used to endorse or promote products derived from this software
48: * without specific prior written permission.
49: *
50: * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
51: * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
52: * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
53: * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
54: * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
55: * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
56: * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
57: * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
58: * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
59: * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
60: * SUCH DAMAGE.
61: *
1.113 fvdl 62: * @(#)vfs_syscalls.c 8.42 (Berkeley) 7/31/95
1.31 cgd 63: */
1.173 lukem 64:
65: #include <sys/cdefs.h>
1.389 ! ad 66: __KERNEL_RCSID(0, "$NetBSD: vfs_syscalls.c,v 1.388 2009/02/15 03:52:49 enami Exp $");
1.111 mrg 67:
1.378 ad 68: #ifdef _KERNEL_OPT
1.258 elad 69: #include "opt_fileassoc.h"
1.261 dogcow 70: #include "veriexec.h"
1.378 ad 71: #endif
1.31 cgd 72:
73: #include <sys/param.h>
74: #include <sys/systm.h>
75: #include <sys/namei.h>
76: #include <sys/filedesc.h>
77: #include <sys/kernel.h>
78: #include <sys/file.h>
79: #include <sys/stat.h>
80: #include <sys/vnode.h>
81: #include <sys/mount.h>
82: #include <sys/proc.h>
83: #include <sys/uio.h>
1.248 yamt 84: #include <sys/kmem.h>
1.31 cgd 85: #include <sys/dirent.h>
1.172 simonb 86: #include <sys/sysctl.h>
1.35 cgd 87: #include <sys/syscallargs.h>
1.306 dsl 88: #include <sys/vfs_syscalls.h>
1.192 drochner 89: #include <sys/ktrace.h>
1.251 elad 90: #ifdef FILEASSOC
91: #include <sys/fileassoc.h>
92: #endif /* FILEASSOC */
1.219 blymn 93: #include <sys/verified_exec.h>
1.242 elad 94: #include <sys/kauth.h>
1.346 ad 95: #include <sys/atomic.h>
1.360 ad 96: #include <sys/module.h>
1.380 pooka 97: #include <sys/buf.h>
1.35 cgd 98:
1.148 fvdl 99: #include <miscfs/genfs/genfs.h>
100: #include <miscfs/syncfs/syncfs.h>
1.342 ad 101: #include <miscfs/specfs/specdev.h>
1.181 thorpej 102:
1.232 jmmv 103: #include <nfs/rpcv2.h>
104: #include <nfs/nfsproto.h>
105: #include <nfs/nfs.h>
106: #include <nfs/nfs_var.h>
107:
1.181 thorpej 108: MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct");
1.110 mrg 109:
1.234 christos 110: static int change_dir(struct nameidata *, struct lwp *);
111: static int change_flags(struct vnode *, u_long, struct lwp *);
112: static int change_mode(struct vnode *, int, struct lwp *l);
113: static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int);
1.63 christos 114:
1.205 junyoung 115: void checkdirs(struct vnode *);
1.31 cgd 116:
1.150 fvdl 117: int dovfsusermount = 0;
118:
1.31 cgd 119: /*
120: * Virtual File System System Calls
121: */
122:
123: /*
124: * Mount a file system.
125: */
1.99 thorpej 126:
127: /*
128: * This table is used to maintain compatibility with 4.3BSD
1.378 ad 129: * and NetBSD 0.9 mount syscalls - and possibly other systems.
130: * Note, the order is important!
1.124 thorpej 131: *
1.167 jdolecek 132: * Do not modify this table. It should only contain filesystems
133: * supported by NetBSD 0.9 and 4.3BSD.
1.99 thorpej 134: */
1.167 jdolecek 135: const char * const mountcompatnames[] = {
1.99 thorpej 136: NULL, /* 0 = MOUNT_NONE */
1.167 jdolecek 137: MOUNT_FFS, /* 1 = MOUNT_UFS */
1.99 thorpej 138: MOUNT_NFS, /* 2 */
139: MOUNT_MFS, /* 3 */
140: MOUNT_MSDOS, /* 4 */
1.167 jdolecek 141: MOUNT_CD9660, /* 5 = MOUNT_ISOFS */
142: MOUNT_FDESC, /* 6 */
143: MOUNT_KERNFS, /* 7 */
144: NULL, /* 8 = MOUNT_DEVFS */
145: MOUNT_AFS, /* 9 */
1.99 thorpej 146: };
1.113 fvdl 147: const int nmountcompatnames = sizeof(mountcompatnames) /
1.99 thorpej 148: sizeof(mountcompatnames[0]);
149:
1.285 elad 150: static int
1.283 elad 151: mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags,
1.324 pooka 152: void *data, size_t *data_len)
1.56 thorpej 153: {
1.113 fvdl 154: struct mount *mp;
1.283 elad 155: int error = 0, saved_flags;
156:
157: mp = vp->v_mount;
158: saved_flags = mp->mnt_flag;
159:
1.329 ad 160: /* We can operate only on VV_ROOT nodes. */
161: if ((vp->v_vflag & VV_ROOT) == 0) {
162: error = EINVAL;
163: goto out;
164: }
1.31 cgd 165:
1.218 yamt 166: /*
1.283 elad 167: * We only allow the filesystem to be reloaded if it
1.373 ad 168: * is currently mounted read-only. Additionally, we
169: * prevent read-write to read-only downgrades.
1.283 elad 170: */
1.373 ad 171: if ((flags & (MNT_RELOAD | MNT_RDONLY)) != 0 &&
172: (mp->mnt_flag & MNT_RDONLY) == 0) {
1.329 ad 173: error = EOPNOTSUPP; /* Needs translation */
174: goto out;
175: }
1.292 elad 176:
177: error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
178: KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data);
179: if (error)
1.329 ad 180: goto out;
1.292 elad 181:
1.358 ad 182: if (vfs_busy(mp, NULL)) {
1.329 ad 183: error = EPERM;
184: goto out;
185: }
1.283 elad 186:
1.358 ad 187: mutex_enter(&mp->mnt_updating);
188:
1.286 yamt 189: mp->mnt_flag &= ~MNT_OP_FLAGS;
190: mp->mnt_flag |= flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE);
191:
1.283 elad 192: /*
193: * Set the mount level flags.
194: */
195: if (flags & MNT_RDONLY)
196: mp->mnt_flag |= MNT_RDONLY;
197: else if (mp->mnt_flag & MNT_RDONLY)
198: mp->mnt_iflag |= IMNT_WANTRDWR;
199: mp->mnt_flag &=
200: ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
201: MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
1.370 simonb 202: MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP |
203: MNT_LOG);
1.283 elad 204: mp->mnt_flag |= flags &
205: (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
206: MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
207: MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP |
1.370 simonb 208: MNT_LOG | MNT_IGNORE);
1.283 elad 209:
1.332 pooka 210: error = VFS_MOUNT(mp, path, data, data_len);
1.283 elad 211:
1.320 dsl 212: if (error && data != NULL) {
1.283 elad 213: int error2;
214:
1.378 ad 215: /*
216: * Update failed; let's try and see if it was an
1.379 ad 217: * export request. For compat with 3.0 and earlier.
1.378 ad 218: */
1.381 ad 219: error2 = vfs_hooks_reexport(mp, path, data);
1.283 elad 220:
1.381 ad 221: /*
222: * Only update error code if the export request was
1.283 elad 223: * understood but some problem occurred while
1.381 ad 224: * processing it.
225: */
1.283 elad 226: if (error2 != EJUSTRETURN)
227: error = error2;
228: }
1.381 ad 229:
1.283 elad 230: if (mp->mnt_iflag & IMNT_WANTRDWR)
231: mp->mnt_flag &= ~MNT_RDONLY;
232: if (error)
233: mp->mnt_flag = saved_flags;
1.286 yamt 234: mp->mnt_flag &= ~MNT_OP_FLAGS;
235: mp->mnt_iflag &= ~IMNT_WANTRDWR;
1.283 elad 236: if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) {
237: if (mp->mnt_syncer == NULL)
238: error = vfs_allocate_syncvnode(mp);
1.125 tls 239: } else {
1.283 elad 240: if (mp->mnt_syncer != NULL)
241: vfs_deallocate_syncvnode(mp);
242: }
1.358 ad 243: mutex_exit(&mp->mnt_updating);
1.354 ad 244: vfs_unbusy(mp, false, NULL);
1.283 elad 245:
1.329 ad 246: out:
1.283 elad 247: return (error);
248: }
249:
1.285 elad 250: static int
1.320 dsl 251: mount_get_vfsops(const char *fstype, struct vfsops **vfsops)
1.283 elad 252: {
1.322 christos 253: char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)];
1.283 elad 254: int error;
255:
1.320 dsl 256: /* Copy file-system type from userspace. */
1.322 christos 257: error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL);
1.63 christos 258: if (error) {
1.54 cgd 259: /*
1.227 jmmv 260: * Historically, filesystem types were identified by numbers.
1.54 cgd 261: * If we get an integer for the filesystem type instead of a
262: * string, we check to see if it matches one of the historic
263: * filesystem types.
1.205 junyoung 264: */
1.283 elad 265: u_long fsindex = (u_long)fstype;
1.99 thorpej 266: if (fsindex >= nmountcompatnames ||
1.320 dsl 267: mountcompatnames[fsindex] == NULL)
268: return ENODEV;
1.331 pooka 269: strlcpy(fstypename, mountcompatnames[fsindex],
270: sizeof(fstypename));
1.31 cgd 271: }
1.283 elad 272:
1.378 ad 273: /* Accept `ufs' as an alias for `ffs', for compatibility. */
1.320 dsl 274: if (strcmp(fstypename, "ufs") == 0)
275: fstypename[0] = 'f';
1.285 elad 276:
1.360 ad 277: if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL)
278: return 0;
279:
280: /* If we can autoload a vfs module, try again */
1.376 ad 281: mutex_enter(&module_lock);
282: (void)module_autoload(fstype, MODULE_CLASS_VFS);
283: mutex_exit(&module_lock);
1.360 ad 284:
285: if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL)
286: return 0;
287:
288: return ENODEV;
1.320 dsl 289: }
290:
291: static int
292: mount_domount(struct lwp *l, struct vnode **vpp, struct vfsops *vfsops,
1.337 ad 293: const char *path, int flags, void *data, size_t *data_len, u_int recurse)
1.320 dsl 294: {
1.366 simonb 295: struct mount *mp;
1.320 dsl 296: struct vnode *vp = *vpp;
297: struct vattr va;
298: int error;
299:
300: error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
301: KAUTH_REQ_SYSTEM_MOUNT_NEW, vp, KAUTH_ARG(flags), data);
302: if (error)
303: return error;
304:
305: /* Can't make a non-dir a mount-point (from here anyway). */
306: if (vp->v_type != VDIR)
307: return ENOTDIR;
308:
309: /*
310: * If the user is not root, ensure that they own the directory
311: * onto which we are attempting to mount.
312: */
1.332 pooka 313: if ((error = VOP_GETATTR(vp, &va, l->l_cred)) != 0 ||
1.320 dsl 314: (va.va_uid != kauth_cred_geteuid(l->l_cred) &&
315: (error = kauth_authorize_generic(l->l_cred,
316: KAUTH_GENERIC_ISSUSER, NULL)) != 0)) {
317: return error;
1.283 elad 318: }
319:
1.320 dsl 320: if (flags & MNT_EXPORTED)
321: return EINVAL;
322:
323: if ((error = vinvalbuf(vp, V_SAVE, l->l_cred, l, 0, 0)) != 0)
324: return error;
325:
1.283 elad 326: /*
327: * Check if a file-system is not already mounted on this vnode.
328: */
1.320 dsl 329: if (vp->v_mountedhere != NULL)
330: return EBUSY;
1.283 elad 331:
1.345 ad 332: mp = kmem_zalloc(sizeof(*mp), KM_SLEEP);
333: if (mp == NULL)
334: return ENOMEM;
1.283 elad 335:
1.320 dsl 336: mp->mnt_op = vfsops;
1.345 ad 337: mp->mnt_refcnt = 1;
1.31 cgd 338:
1.296 pooka 339: TAILQ_INIT(&mp->mnt_vnodelist);
1.358 ad 340: rw_init(&mp->mnt_unmounting);
1.344 dholland 341: mutex_init(&mp->mnt_renamelock, MUTEX_DEFAULT, IPL_NONE);
1.358 ad 342: mutex_init(&mp->mnt_updating, MUTEX_DEFAULT, IPL_NONE);
343: error = vfs_busy(mp, NULL);
344: KASSERT(error == 0);
345: mutex_enter(&mp->mnt_updating);
1.283 elad 346:
1.129 fvdl 347: mp->mnt_vnodecovered = vp;
1.257 ad 348: mp->mnt_stat.f_owner = kauth_cred_geteuid(l->l_cred);
1.277 hannken 349: mount_initspecific(mp);
1.195 thorpej 350:
351: /*
352: * The underlying file system may refuse the mount for
1.198 thorpej 353: * various reasons. Allow the user to force it to happen.
1.286 yamt 354: *
1.284 elad 355: * Set the mount level flags.
356: */
1.286 yamt 357: mp->mnt_flag = flags &
358: (MNT_FORCE | MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
1.284 elad 359: MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
360: MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP |
1.370 simonb 361: MNT_LOG | MNT_IGNORE | MNT_RDONLY);
1.284 elad 362:
1.332 pooka 363: error = VFS_MOUNT(mp, path, data, data_len);
1.286 yamt 364: mp->mnt_flag &= ~MNT_OP_FLAGS;
1.232 jmmv 365:
1.31 cgd 366: /*
367: * Put the new filesystem on the mount list after root.
368: */
369: cache_purge(vp);
1.320 dsl 370: if (error != 0) {
1.283 elad 371: vp->v_mountedhere = NULL;
1.358 ad 372: mutex_exit(&mp->mnt_updating);
1.354 ad 373: vfs_unbusy(mp, false, NULL);
1.358 ad 374: vfs_destroy(mp);
1.320 dsl 375: return error;
1.31 cgd 376: }
1.283 elad 377:
1.320 dsl 378: mp->mnt_iflag &= ~IMNT_WANTRDWR;
1.345 ad 379: mutex_enter(&mountlist_lock);
1.320 dsl 380: vp->v_mountedhere = mp;
381: CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
1.329 ad 382: mutex_exit(&mountlist_lock);
1.337 ad 383: vn_restorerecurse(vp, recurse);
1.320 dsl 384: VOP_UNLOCK(vp, 0);
385: checkdirs(vp);
386: if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
387: error = vfs_allocate_syncvnode(mp);
1.347 ad 388: /* Hold an additional reference to the mount across VFS_START(). */
1.358 ad 389: mutex_exit(&mp->mnt_updating);
1.354 ad 390: vfs_unbusy(mp, true, NULL);
1.332 pooka 391: (void) VFS_STATVFS(mp, &mp->mnt_stat);
392: error = VFS_START(mp, 0);
1.372 ad 393: if (error)
1.320 dsl 394: vrele(vp);
1.347 ad 395: /* Drop reference held for VFS_START(). */
1.358 ad 396: vfs_destroy(mp);
1.320 dsl 397: *vpp = NULL;
398: return error;
1.283 elad 399: }
400:
401: static int
402: mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags,
1.324 pooka 403: void *data, size_t *data_len)
1.283 elad 404: {
405: struct mount *mp;
406: int error;
407:
1.289 elad 408: /* If MNT_GETARGS is specified, it should be the only flag. */
1.320 dsl 409: if (flags & ~MNT_GETARGS)
410: return EINVAL;
1.289 elad 411:
1.283 elad 412: mp = vp->v_mount;
413:
1.292 elad 414: /* XXX: probably some notion of "can see" here if we want isolation. */
415: error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
416: KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL);
417: if (error)
1.320 dsl 418: return error;
1.292 elad 419:
1.329 ad 420: if ((vp->v_vflag & VV_ROOT) == 0)
1.320 dsl 421: return EINVAL;
1.283 elad 422:
1.358 ad 423: if (vfs_busy(mp, NULL))
1.320 dsl 424: return EPERM;
1.283 elad 425:
1.358 ad 426: mutex_enter(&mp->mnt_updating);
1.286 yamt 427: mp->mnt_flag &= ~MNT_OP_FLAGS;
428: mp->mnt_flag |= MNT_GETARGS;
1.332 pooka 429: error = VFS_MOUNT(mp, path, data, data_len);
1.286 yamt 430: mp->mnt_flag &= ~MNT_OP_FLAGS;
1.358 ad 431: mutex_exit(&mp->mnt_updating);
1.283 elad 432:
1.354 ad 433: vfs_unbusy(mp, false, NULL);
1.283 elad 434: return (error);
435: }
436:
1.321 dsl 437: int
1.335 dsl 438: sys___mount50(struct lwp *l, const struct sys___mount50_args *uap, register_t *retval)
1.321 dsl 439: {
1.335 dsl 440: /* {
1.321 dsl 441: syscallarg(const char *) type;
442: syscallarg(const char *) path;
443: syscallarg(int) flags;
444: syscallarg(void *) data;
445: syscallarg(size_t) data_len;
1.335 dsl 446: } */
1.321 dsl 447:
448: return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path),
449: SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE,
450: SCARG(uap, data_len), retval);
451: }
1.320 dsl 452:
453: int
454: do_sys_mount(struct lwp *l, struct vfsops *vfsops, const char *type,
455: const char *path, int flags, void *data, enum uio_seg data_seg,
456: size_t data_len, register_t *retval)
457: {
1.283 elad 458: struct vnode *vp;
459: struct nameidata nd;
1.320 dsl 460: void *data_buf = data;
1.337 ad 461: u_int recurse;
1.283 elad 462: int error;
463:
464: /*
465: * Get vnode to be covered
466: */
1.334 pooka 467: NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path);
1.283 elad 468: if ((error = namei(&nd)) != 0)
469: return (error);
470: vp = nd.ni_vp;
471:
472: /*
473: * A lookup in VFS_MOUNT might result in an attempt to
474: * lock this vnode again, so make the lock recursive.
475: */
1.320 dsl 476: if (vfsops == NULL) {
1.361 ad 477: if (flags & (MNT_GETARGS | MNT_UPDATE)) {
478: vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
479: recurse = vn_setrecurse(vp);
1.320 dsl 480: vfsops = vp->v_mount->mnt_op;
1.361 ad 481: } else {
1.320 dsl 482: /* 'type' is userspace */
483: error = mount_get_vfsops(type, &vfsops);
1.361 ad 484: vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
485: recurse = vn_setrecurse(vp);
1.320 dsl 486: if (error != 0)
487: goto done;
488: }
1.361 ad 489: } else {
490: vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
491: recurse = vn_setrecurse(vp);
1.320 dsl 492: }
493:
494: if (data != NULL && data_seg == UIO_USERSPACE) {
495: if (data_len == 0) {
496: /* No length supplied, use default for filesystem */
497: data_len = vfsops->vfs_min_mount_data;
498: if (data_len > VFS_MAX_MOUNT_DATA) {
499: error = EINVAL;
500: goto done;
501: }
1.378 ad 502: /*
503: * Hopefully a longer buffer won't make copyin() fail.
504: * For compatibility with 3.0 and earlier.
505: */
1.320 dsl 506: if (flags & MNT_UPDATE
507: && data_len < sizeof (struct mnt_export_args30))
508: data_len = sizeof (struct mnt_export_args30);
509: }
1.384 yamt 510: data_buf = kmem_alloc(data_len, KM_SLEEP);
1.283 elad 511:
1.320 dsl 512: /* NFS needs the buffer even for mnt_getargs .... */
513: error = copyin(data, data_buf, data_len);
514: if (error != 0)
515: goto done;
516: }
517:
518: if (flags & MNT_GETARGS) {
519: if (data_len == 0) {
520: error = EINVAL;
521: goto done;
522: }
1.324 pooka 523: error = mount_getargs(l, vp, path, flags, data_buf, &data_len);
1.320 dsl 524: if (error != 0)
525: goto done;
526: if (data_seg == UIO_USERSPACE)
527: error = copyout(data_buf, data, data_len);
528: *retval = data_len;
529: } else if (flags & MNT_UPDATE) {
1.324 pooka 530: error = mount_update(l, vp, path, flags, data_buf, &data_len);
1.283 elad 531: } else {
532: /* Locking is handled internally in mount_domount(). */
1.320 dsl 533: error = mount_domount(l, &vp, vfsops, path, flags, data_buf,
1.337 ad 534: &data_len, recurse);
1.283 elad 535: }
536:
1.320 dsl 537: done:
1.337 ad 538: if (vp != NULL) {
539: vn_restorerecurse(vp, recurse);
540: vput(vp);
541: }
1.320 dsl 542: if (data_buf != data)
1.384 yamt 543: kmem_free(data_buf, data_len);
1.31 cgd 544: return (error);
545: }
546:
547: /*
1.43 mycroft 548: * Scan all active processes to see if any of them have a current
549: * or root directory onto which the new filesystem has just been
550: * mounted. If so, replace them with the new mount point.
551: */
1.63 christos 552: void
1.221 thorpej 553: checkdirs(struct vnode *olddp)
1.43 mycroft 554: {
1.134 thorpej 555: struct cwdinfo *cwdi;
1.355 ad 556: struct vnode *newdp, *rele1, *rele2;
1.43 mycroft 557: struct proc *p;
1.355 ad 558: bool retry;
1.43 mycroft 559:
560: if (olddp->v_usecount == 1)
561: return;
1.189 thorpej 562: if (VFS_ROOT(olddp->v_mountedhere, &newdp))
1.43 mycroft 563: panic("mount: lost mount");
1.355 ad 564:
565: do {
566: retry = false;
567: mutex_enter(proc_lock);
568: PROCLIST_FOREACH(p, &allproc) {
569: if ((p->p_flag & PK_MARKER) != 0)
570: continue;
571: if ((cwdi = p->p_cwdi) == NULL)
572: continue;
573: /*
574: * Can't change to the old directory any more,
575: * so even if we see a stale value it's not a
576: * problem.
577: */
578: if (cwdi->cwdi_cdir != olddp &&
579: cwdi->cwdi_rdir != olddp)
580: continue;
581: retry = true;
582: rele1 = NULL;
583: rele2 = NULL;
584: atomic_inc_uint(&cwdi->cwdi_refcnt);
585: mutex_exit(proc_lock);
586: rw_enter(&cwdi->cwdi_lock, RW_WRITER);
587: if (cwdi->cwdi_cdir == olddp) {
588: rele1 = cwdi->cwdi_cdir;
589: VREF(newdp);
590: cwdi->cwdi_cdir = newdp;
591: }
592: if (cwdi->cwdi_rdir == olddp) {
593: rele2 = cwdi->cwdi_rdir;
594: VREF(newdp);
595: cwdi->cwdi_rdir = newdp;
596: }
597: rw_exit(&cwdi->cwdi_lock);
598: cwdfree(cwdi);
599: if (rele1 != NULL)
600: vrele(rele1);
601: if (rele2 != NULL)
602: vrele(rele2);
603: mutex_enter(proc_lock);
604: break;
1.43 mycroft 605: }
1.355 ad 606: mutex_exit(proc_lock);
607: } while (retry);
608:
1.43 mycroft 609: if (rootvnode == olddp) {
610: vrele(rootvnode);
611: VREF(newdp);
612: rootvnode = newdp;
613: }
614: vput(newdp);
615: }
616:
617: /*
1.31 cgd 618: * Unmount a file system.
619: *
620: * Note: unmount takes a path to the vnode mounted on as argument,
621: * not special file (as before).
622: */
623: /* ARGSUSED */
1.63 christos 624: int
1.335 dsl 625: sys_unmount(struct lwp *l, const struct sys_unmount_args *uap, register_t *retval)
1.56 thorpej 626: {
1.335 dsl 627: /* {
1.74 cgd 628: syscallarg(const char *) path;
1.35 cgd 629: syscallarg(int) flags;
1.335 dsl 630: } */
1.155 augustss 631: struct vnode *vp;
1.31 cgd 632: struct mount *mp;
633: int error;
634: struct nameidata nd;
635:
1.310 dsl 636: NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
1.334 pooka 637: SCARG(uap, path));
1.63 christos 638: if ((error = namei(&nd)) != 0)
1.31 cgd 639: return (error);
640: vp = nd.ni_vp;
1.43 mycroft 641: mp = vp->v_mount;
1.358 ad 642: atomic_inc_uint(&mp->mnt_refcnt);
1.345 ad 643: VOP_UNLOCK(vp, 0);
1.31 cgd 644:
1.295 elad 645: error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
646: KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL);
647: if (error) {
1.345 ad 648: vrele(vp);
1.358 ad 649: vfs_destroy(mp);
1.31 cgd 650: return (error);
651: }
652:
653: /*
1.47 mycroft 654: * Don't allow unmounting the root file system.
655: */
656: if (mp->mnt_flag & MNT_ROOTFS) {
1.345 ad 657: vrele(vp);
1.358 ad 658: vfs_destroy(mp);
1.47 mycroft 659: return (EINVAL);
660: }
661:
662: /*
1.31 cgd 663: * Must be the root of the filesystem
664: */
1.329 ad 665: if ((vp->v_vflag & VV_ROOT) == 0) {
1.345 ad 666: vrele(vp);
1.358 ad 667: vfs_destroy(mp);
1.31 cgd 668: return (EINVAL);
669: }
1.78 fvdl 670:
1.359 ad 671: vrele(vp);
1.358 ad 672: error = dounmount(mp, SCARG(uap, flags), l);
1.374 ad 673: vfs_destroy(mp);
1.358 ad 674: return error;
1.31 cgd 675: }
676:
677: /*
1.358 ad 678: * Do the actual file system unmount. File system is assumed to have
679: * been locked by the caller.
680: *
1.375 wiz 681: * => Caller hold reference to the mount, explicitly for dounmount().
1.31 cgd 682: */
1.63 christos 683: int
1.234 christos 684: dounmount(struct mount *mp, int flags, struct lwp *l)
1.31 cgd 685: {
686: struct vnode *coveredvp;
687: int error;
1.140 sommerfe 688: int async;
1.162 fvdl 689: int used_syncer;
1.31 cgd 690:
1.256 elad 691: #if NVERIEXEC > 0
1.279 elad 692: error = veriexec_unmountchk(mp);
693: if (error)
694: return (error);
1.256 elad 695: #endif /* NVERIEXEC > 0 */
1.251 elad 696:
1.358 ad 697: /*
698: * XXX Freeze syncer. Must do this before locking the
699: * mount point. See dounmount() for details.
700: */
701: mutex_enter(&syncer_mutex);
702: rw_enter(&mp->mnt_unmounting, RW_WRITER);
703: if ((mp->mnt_iflag & IMNT_GONE) != 0) {
704: rw_exit(&mp->mnt_unmounting);
705: mutex_exit(&syncer_mutex);
706: return ENOENT;
707: }
708:
1.162 fvdl 709: used_syncer = (mp->mnt_syncer != NULL);
710:
1.148 fvdl 711: /*
1.165 thorpej 712: * XXX Syncer must be frozen when we get here. This should really
713: * be done on a per-mountpoint basis, but especially the softdep
1.227 jmmv 714: * code possibly called from the syncer doesn't exactly work on a
1.165 thorpej 715: * per-mountpoint basis, so the softdep code would become a maze
716: * of vfs_busy() calls.
717: *
1.300 ad 718: * The caller of dounmount() must acquire syncer_mutex because
719: * the syncer itself acquires locks in syncer_mutex -> vfs_busy
1.165 thorpej 720: * order, and we must preserve that order to avoid deadlock.
721: *
722: * So, if the file system did not use the syncer, now is
1.300 ad 723: * the time to release the syncer_mutex.
1.148 fvdl 724: */
1.165 thorpej 725: if (used_syncer == 0)
1.300 ad 726: mutex_exit(&syncer_mutex);
1.148 fvdl 727:
1.196 dbj 728: mp->mnt_iflag |= IMNT_UNMOUNT;
1.141 sommerfe 729: async = mp->mnt_flag & MNT_ASYNC;
1.151 mycroft 730: mp->mnt_flag &= ~MNT_ASYNC;
1.31 cgd 731: cache_purgevfs(mp); /* remove cache entries for this file sys */
1.160 mycroft 732: if (mp->mnt_syncer != NULL)
733: vfs_deallocate_syncvnode(mp);
1.209 hannken 734: error = 0;
735: if ((mp->mnt_flag & MNT_RDONLY) == 0) {
1.371 hannken 736: error = VFS_SYNC(mp, MNT_WAIT, l->l_cred);
1.209 hannken 737: }
1.342 ad 738: vfs_scrubvnlist(mp);
1.209 hannken 739: if (error == 0 || (flags & MNT_FORCE))
1.332 pooka 740: error = VFS_UNMOUNT(mp, flags);
1.31 cgd 741: if (error) {
1.151 mycroft 742: if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
1.148 fvdl 743: (void) vfs_allocate_syncvnode(mp);
1.196 dbj 744: mp->mnt_iflag &= ~IMNT_UNMOUNT;
1.140 sommerfe 745: mp->mnt_flag |= async;
1.358 ad 746: rw_exit(&mp->mnt_unmounting);
1.162 fvdl 747: if (used_syncer)
1.300 ad 748: mutex_exit(&syncer_mutex);
1.142 sommerfe 749: return (error);
1.113 fvdl 750: }
1.338 ad 751: vfs_scrubvnlist(mp);
1.358 ad 752: mutex_enter(&mountlist_lock);
1.297 hannken 753: if ((coveredvp = mp->mnt_vnodecovered) != NULLVP)
1.113 fvdl 754: coveredvp->v_mountedhere = NULL;
1.358 ad 755: CIRCLEQ_REMOVE(&mountlist, mp, mnt_list);
756: mp->mnt_iflag |= IMNT_GONE;
757: mutex_exit(&mountlist_lock);
1.273 reinoud 758: if (TAILQ_FIRST(&mp->mnt_vnodelist) != NULL)
1.113 fvdl 759: panic("unmount: dangling vnode");
1.162 fvdl 760: if (used_syncer)
1.300 ad 761: mutex_exit(&syncer_mutex);
1.231 jmmv 762: vfs_hooks_unmount(mp);
1.358 ad 763: rw_exit(&mp->mnt_unmounting);
1.374 ad 764: vfs_destroy(mp); /* reference from mount() */
1.345 ad 765: if (coveredvp != NULLVP)
766: vrele(coveredvp);
1.113 fvdl 767: return (0);
1.31 cgd 768: }
769:
770: /*
771: * Sync each mounted filesystem.
772: */
773: #ifdef DEBUG
774: int syncprt = 0;
775: struct ctldebug debug0 = { "syncprt", &syncprt };
776: #endif
777:
778: /* ARGSUSED */
1.63 christos 779: int
1.335 dsl 780: sys_sync(struct lwp *l, const void *v, register_t *retval)
1.31 cgd 781: {
1.155 augustss 782: struct mount *mp, *nmp;
1.31 cgd 783: int asyncflag;
1.257 ad 784:
785: if (l == NULL)
786: l = &lwp0;
1.31 cgd 787:
1.329 ad 788: mutex_enter(&mountlist_lock);
1.353 ad 789: for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist;
790: mp = nmp) {
1.358 ad 791: if (vfs_busy(mp, &nmp)) {
1.113 fvdl 792: continue;
793: }
1.358 ad 794: mutex_enter(&mp->mnt_updating);
1.307 hannken 795: if ((mp->mnt_flag & MNT_RDONLY) == 0) {
1.31 cgd 796: asyncflag = mp->mnt_flag & MNT_ASYNC;
797: mp->mnt_flag &= ~MNT_ASYNC;
1.332 pooka 798: VFS_SYNC(mp, MNT_NOWAIT, l->l_cred);
1.31 cgd 799: if (asyncflag)
1.113 fvdl 800: mp->mnt_flag |= MNT_ASYNC;
1.31 cgd 801: }
1.358 ad 802: mutex_exit(&mp->mnt_updating);
1.354 ad 803: vfs_unbusy(mp, false, &nmp);
1.31 cgd 804: }
1.329 ad 805: mutex_exit(&mountlist_lock);
1.31 cgd 806: #ifdef DEBUG
807: if (syncprt)
808: vfs_bufstats();
809: #endif /* DEBUG */
810: return (0);
811: }
812:
813: /*
814: * Change filesystem quotas.
815: */
816: /* ARGSUSED */
1.63 christos 817: int
1.335 dsl 818: sys_quotactl(struct lwp *l, const struct sys_quotactl_args *uap, register_t *retval)
1.56 thorpej 819: {
1.335 dsl 820: /* {
1.74 cgd 821: syscallarg(const char *) path;
1.35 cgd 822: syscallarg(int) cmd;
823: syscallarg(int) uid;
1.272 christos 824: syscallarg(void *) arg;
1.335 dsl 825: } */
1.155 augustss 826: struct mount *mp;
1.31 cgd 827: int error;
828: struct nameidata nd;
829:
1.326 pooka 830: NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
1.334 pooka 831: SCARG(uap, path));
1.63 christos 832: if ((error = namei(&nd)) != 0)
1.31 cgd 833: return (error);
1.307 hannken 834: mp = nd.ni_vp->v_mount;
1.197 hannken 835: error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid),
1.332 pooka 836: SCARG(uap, arg));
1.326 pooka 837: vrele(nd.ni_vp);
1.197 hannken 838: return (error);
1.31 cgd 839: }
840:
1.206 christos 841: int
1.234 christos 842: dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags,
1.185 christos 843: int root)
844: {
1.234 christos 845: struct cwdinfo *cwdi = l->l_proc->p_cwdi;
1.185 christos 846: int error = 0;
847:
848: /*
849: * If MNT_NOWAIT or MNT_LAZY is specified, do not
1.204 dbj 850: * refresh the fsstat cache. MNT_WAIT or MNT_LAZY
1.185 christos 851: * overrides MNT_NOWAIT.
852: */
853: if (flags == MNT_NOWAIT || flags == MNT_LAZY ||
854: (flags != MNT_WAIT && flags != 0)) {
855: memcpy(sp, &mp->mnt_stat, sizeof(*sp));
856: goto done;
857: }
1.205 junyoung 858:
1.211 jdolecek 859: /* Get the filesystem stats now */
860: memset(sp, 0, sizeof(*sp));
1.332 pooka 861: if ((error = VFS_STATVFS(mp, sp)) != 0) {
1.185 christos 862: return error;
863: }
864:
865: if (cwdi->cwdi_rdir == NULL)
866: (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat));
867: done:
868: if (cwdi->cwdi_rdir != NULL) {
869: size_t len;
870: char *bp;
1.364 christos 871: char c;
1.236 yamt 872: char *path = PNBUF_GET();
1.185 christos 873:
874: bp = path + MAXPATHLEN;
875: *--bp = '\0';
1.328 ad 876: rw_enter(&cwdi->cwdi_lock, RW_READER);
1.185 christos 877: error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path,
1.234 christos 878: MAXPATHLEN / 2, 0, l);
1.328 ad 879: rw_exit(&cwdi->cwdi_lock);
1.185 christos 880: if (error) {
1.236 yamt 881: PNBUF_PUT(path);
1.185 christos 882: return error;
883: }
884: len = strlen(bp);
1.387 christos 885: if (len != 1) {
886: /*
887: * for mount points that are below our root, we can see
888: * them, so we fix up the pathname and return them. The
889: * rest we cannot see, so we don't allow viewing the
890: * data.
891: */
892: if (strncmp(bp, sp->f_mntonname, len) == 0 &&
893: ((c = sp->f_mntonname[len]) == '/' || c == '\0')) {
894: (void)strlcpy(sp->f_mntonname,
1.388 enami 895: c == '\0' ? "/" : &sp->f_mntonname[len],
1.187 itojun 896: sizeof(sp->f_mntonname));
1.387 christos 897: } else {
898: if (root)
899: (void)strlcpy(sp->f_mntonname, "/",
900: sizeof(sp->f_mntonname));
901: else
902: error = EPERM;
903: }
1.185 christos 904: }
1.236 yamt 905: PNBUF_PUT(path);
1.185 christos 906: }
1.206 christos 907: sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK;
1.185 christos 908: return error;
909: }
910:
1.31 cgd 911: /*
1.311 dsl 912: * Get filesystem statistics by path.
1.31 cgd 913: */
1.311 dsl 914: int
915: do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb)
916: {
917: struct mount *mp;
918: int error;
919: struct nameidata nd;
920:
1.334 pooka 921: NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path);
1.311 dsl 922: if ((error = namei(&nd)) != 0)
923: return error;
924: mp = nd.ni_vp->v_mount;
925: error = dostatvfs(mp, sb, l, flags, 1);
926: vrele(nd.ni_vp);
927: return error;
928: }
929:
1.31 cgd 930: /* ARGSUSED */
1.63 christos 931: int
1.335 dsl 932: sys_statvfs1(struct lwp *l, const struct sys_statvfs1_args *uap, register_t *retval)
1.56 thorpej 933: {
1.335 dsl 934: /* {
1.74 cgd 935: syscallarg(const char *) path;
1.206 christos 936: syscallarg(struct statvfs *) buf;
937: syscallarg(int) flags;
1.335 dsl 938: } */
1.241 yamt 939: struct statvfs *sb;
1.31 cgd 940: int error;
941:
1.241 yamt 942: sb = STATVFSBUF_GET();
1.311 dsl 943: error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb);
944: if (error == 0)
1.241 yamt 945: error = copyout(sb, SCARG(uap, buf), sizeof(*sb));
946: STATVFSBUF_PUT(sb);
947: return error;
1.31 cgd 948: }
949:
950: /*
1.311 dsl 951: * Get filesystem statistics by fd.
1.31 cgd 952: */
1.311 dsl 953: int
954: do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb)
955: {
1.346 ad 956: file_t *fp;
1.311 dsl 957: struct mount *mp;
958: int error;
959:
1.346 ad 960: /* fd_getvnode() will use the descriptor for us */
961: if ((error = fd_getvnode(fd, &fp)) != 0)
1.311 dsl 962: return (error);
963: mp = ((struct vnode *)fp->f_data)->v_mount;
1.346 ad 964: error = dostatvfs(mp, sb, curlwp, flags, 1);
965: fd_putfile(fd);
1.311 dsl 966: return error;
967: }
968:
1.31 cgd 969: /* ARGSUSED */
1.63 christos 970: int
1.335 dsl 971: sys_fstatvfs1(struct lwp *l, const struct sys_fstatvfs1_args *uap, register_t *retval)
1.56 thorpej 972: {
1.335 dsl 973: /* {
1.35 cgd 974: syscallarg(int) fd;
1.206 christos 975: syscallarg(struct statvfs *) buf;
976: syscallarg(int) flags;
1.335 dsl 977: } */
1.241 yamt 978: struct statvfs *sb;
1.31 cgd 979: int error;
980:
1.241 yamt 981: sb = STATVFSBUF_GET();
1.311 dsl 982: error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb);
1.316 dsl 983: if (error == 0)
1.311 dsl 984: error = copyout(sb, SCARG(uap, buf), sizeof(*sb));
1.241 yamt 985: STATVFSBUF_PUT(sb);
1.185 christos 986: return error;
1.31 cgd 987: }
988:
1.185 christos 989:
1.31 cgd 990: /*
991: * Get statistics on all filesystems.
992: */
1.63 christos 993: int
1.311 dsl 994: do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags,
995: int (*copyfn)(const void *, void *, size_t), size_t entry_sz,
996: register_t *retval)
1.56 thorpej 997: {
1.185 christos 998: int root = 0;
1.179 thorpej 999: struct proc *p = l->l_proc;
1.155 augustss 1000: struct mount *mp, *nmp;
1.241 yamt 1001: struct statvfs *sb;
1.206 christos 1002: size_t count, maxcount;
1003: int error = 0;
1.31 cgd 1004:
1.241 yamt 1005: sb = STATVFSBUF_GET();
1.311 dsl 1006: maxcount = bufsize / entry_sz;
1.329 ad 1007: mutex_enter(&mountlist_lock);
1.113 fvdl 1008: count = 0;
1.176 matt 1009: for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist;
1010: mp = nmp) {
1.358 ad 1011: if (vfs_busy(mp, &nmp)) {
1.113 fvdl 1012: continue;
1013: }
1014: if (sfsp && count < maxcount) {
1.311 dsl 1015: error = dostatvfs(mp, sb, l, flags, 0);
1.185 christos 1016: if (error) {
1.354 ad 1017: vfs_unbusy(mp, false, &nmp);
1.365 christos 1018: error = 0;
1.31 cgd 1019: continue;
1.113 fvdl 1020: }
1.311 dsl 1021: error = copyfn(sb, sfsp, entry_sz);
1.133 mycroft 1022: if (error) {
1.354 ad 1023: vfs_unbusy(mp, false, NULL);
1.241 yamt 1024: goto out;
1.133 mycroft 1025: }
1.311 dsl 1026: sfsp = (char *)sfsp + entry_sz;
1.241 yamt 1027: root |= strcmp(sb->f_mntonname, "/") == 0;
1.31 cgd 1028: }
1029: count++;
1.354 ad 1030: vfs_unbusy(mp, false, &nmp);
1.31 cgd 1031: }
1.358 ad 1032: mutex_exit(&mountlist_lock);
1.331 pooka 1033:
1.185 christos 1034: if (root == 0 && p->p_cwdi->cwdi_rdir) {
1035: /*
1036: * fake a root entry
1037: */
1.331 pooka 1038: error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount,
1039: sb, l, flags, 1);
1.311 dsl 1040: if (error != 0)
1.241 yamt 1041: goto out;
1.365 christos 1042: if (sfsp) {
1.311 dsl 1043: error = copyfn(sb, sfsp, entry_sz);
1.365 christos 1044: if (error != 0)
1045: goto out;
1046: }
1.185 christos 1047: count++;
1048: }
1.31 cgd 1049: if (sfsp && count > maxcount)
1050: *retval = maxcount;
1051: else
1052: *retval = count;
1.241 yamt 1053: out:
1054: STATVFSBUF_PUT(sb);
1.185 christos 1055: return error;
1.31 cgd 1056: }
1057:
1.311 dsl 1058: int
1.335 dsl 1059: sys_getvfsstat(struct lwp *l, const struct sys_getvfsstat_args *uap, register_t *retval)
1.311 dsl 1060: {
1.335 dsl 1061: /* {
1.311 dsl 1062: syscallarg(struct statvfs *) buf;
1063: syscallarg(size_t) bufsize;
1064: syscallarg(int) flags;
1.335 dsl 1065: } */
1.311 dsl 1066:
1067: return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize),
1068: SCARG(uap, flags), copyout, sizeof (struct statvfs), retval);
1069: }
1070:
1.31 cgd 1071: /*
1072: * Change current working directory to a given file descriptor.
1073: */
1074: /* ARGSUSED */
1.63 christos 1075: int
1.335 dsl 1076: sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval)
1.56 thorpej 1077: {
1.335 dsl 1078: /* {
1.35 cgd 1079: syscallarg(int) fd;
1.335 dsl 1080: } */
1.179 thorpej 1081: struct proc *p = l->l_proc;
1.328 ad 1082: struct cwdinfo *cwdi;
1.43 mycroft 1083: struct vnode *vp, *tdp;
1084: struct mount *mp;
1.346 ad 1085: file_t *fp;
1086: int error, fd;
1.31 cgd 1087:
1.346 ad 1088: /* fd_getvnode() will use the descriptor for us */
1089: fd = SCARG(uap, fd);
1090: if ((error = fd_getvnode(fd, &fp)) != 0)
1.31 cgd 1091: return (error);
1.346 ad 1092: vp = fp->f_data;
1.131 sommerfe 1093:
1.43 mycroft 1094: VREF(vp);
1.113 fvdl 1095: vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1.31 cgd 1096: if (vp->v_type != VDIR)
1097: error = ENOTDIR;
1098: else
1.332 pooka 1099: error = VOP_ACCESS(vp, VEXEC, l->l_cred);
1.303 pooka 1100: if (error) {
1101: vput(vp);
1102: goto out;
1103: }
1.304 pooka 1104: while ((mp = vp->v_mountedhere) != NULL) {
1.358 ad 1105: error = vfs_busy(mp, NULL);
1.298 chs 1106: vput(vp);
1.357 xtraeme 1107: if (error != 0)
1.356 ad 1108: goto out;
1.189 thorpej 1109: error = VFS_ROOT(mp, &tdp);
1.354 ad 1110: vfs_unbusy(mp, false, NULL);
1.113 fvdl 1111: if (error)
1.303 pooka 1112: goto out;
1.43 mycroft 1113: vp = tdp;
1114: }
1.113 fvdl 1115: VOP_UNLOCK(vp, 0);
1.131 sommerfe 1116:
1117: /*
1118: * Disallow changing to a directory not under the process's
1119: * current root directory (if there is one).
1120: */
1.328 ad 1121: cwdi = p->p_cwdi;
1122: rw_enter(&cwdi->cwdi_lock, RW_WRITER);
1.234 christos 1123: if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) {
1.131 sommerfe 1124: vrele(vp);
1.135 thorpej 1125: error = EPERM; /* operation not permitted */
1.328 ad 1126: } else {
1127: vrele(cwdi->cwdi_cdir);
1128: cwdi->cwdi_cdir = vp;
1.131 sommerfe 1129: }
1.328 ad 1130: rw_exit(&cwdi->cwdi_lock);
1.205 junyoung 1131:
1.135 thorpej 1132: out:
1.346 ad 1133: fd_putfile(fd);
1.135 thorpej 1134: return (error);
1.31 cgd 1135: }
1136:
1137: /*
1.221 thorpej 1138: * Change this process's notion of the root directory to a given file
1139: * descriptor.
1.131 sommerfe 1140: */
1141: int
1.335 dsl 1142: sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval)
1.131 sommerfe 1143: {
1.179 thorpej 1144: struct proc *p = l->l_proc;
1.328 ad 1145: struct cwdinfo *cwdi;
1.131 sommerfe 1146: struct vnode *vp;
1.346 ad 1147: file_t *fp;
1148: int error, fd = SCARG(uap, fd);
1.131 sommerfe 1149:
1.268 elad 1150: if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT,
1151: KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0)
1.131 sommerfe 1152: return error;
1.346 ad 1153: /* fd_getvnode() will use the descriptor for us */
1154: if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
1.131 sommerfe 1155: return error;
1.346 ad 1156: vp = fp->f_data;
1.131 sommerfe 1157: vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1158: if (vp->v_type != VDIR)
1159: error = ENOTDIR;
1160: else
1.332 pooka 1161: error = VOP_ACCESS(vp, VEXEC, l->l_cred);
1.131 sommerfe 1162: VOP_UNLOCK(vp, 0);
1163: if (error)
1.135 thorpej 1164: goto out;
1.131 sommerfe 1165: VREF(vp);
1166:
1167: /*
1168: * Prevent escaping from chroot by putting the root under
1169: * the working directory. Silently chdir to / if we aren't
1170: * already there.
1171: */
1.328 ad 1172: cwdi = p->p_cwdi;
1173: rw_enter(&cwdi->cwdi_lock, RW_WRITER);
1.234 christos 1174: if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) {
1.131 sommerfe 1175: /*
1176: * XXX would be more failsafe to change directory to a
1177: * deadfs node here instead
1178: */
1.134 thorpej 1179: vrele(cwdi->cwdi_cdir);
1.131 sommerfe 1180: VREF(vp);
1.134 thorpej 1181: cwdi->cwdi_cdir = vp;
1.131 sommerfe 1182: }
1.205 junyoung 1183:
1.134 thorpej 1184: if (cwdi->cwdi_rdir != NULL)
1185: vrele(cwdi->cwdi_rdir);
1186: cwdi->cwdi_rdir = vp;
1.328 ad 1187: rw_exit(&cwdi->cwdi_lock);
1188:
1.135 thorpej 1189: out:
1.346 ad 1190: fd_putfile(fd);
1.135 thorpej 1191: return (error);
1.131 sommerfe 1192: }
1193:
1194: /*
1.31 cgd 1195: * Change current working directory (``.'').
1196: */
1197: /* ARGSUSED */
1.63 christos 1198: int
1.335 dsl 1199: sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval)
1.56 thorpej 1200: {
1.335 dsl 1201: /* {
1.74 cgd 1202: syscallarg(const char *) path;
1.335 dsl 1203: } */
1.179 thorpej 1204: struct proc *p = l->l_proc;
1.328 ad 1205: struct cwdinfo *cwdi;
1.31 cgd 1206: int error;
1207: struct nameidata nd;
1208:
1.310 dsl 1209: NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
1.334 pooka 1210: SCARG(uap, path));
1.234 christos 1211: if ((error = change_dir(&nd, l)) != 0)
1.31 cgd 1212: return (error);
1.328 ad 1213: cwdi = p->p_cwdi;
1214: rw_enter(&cwdi->cwdi_lock, RW_WRITER);
1.134 thorpej 1215: vrele(cwdi->cwdi_cdir);
1216: cwdi->cwdi_cdir = nd.ni_vp;
1.328 ad 1217: rw_exit(&cwdi->cwdi_lock);
1.31 cgd 1218: return (0);
1219: }
1220:
1221: /*
1222: * Change notion of root (``/'') directory.
1223: */
1224: /* ARGSUSED */
1.63 christos 1225: int
1.335 dsl 1226: sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval)
1.56 thorpej 1227: {
1.335 dsl 1228: /* {
1.74 cgd 1229: syscallarg(const char *) path;
1.335 dsl 1230: } */
1.179 thorpej 1231: struct proc *p = l->l_proc;
1.328 ad 1232: struct cwdinfo *cwdi;
1.131 sommerfe 1233: struct vnode *vp;
1.31 cgd 1234: int error;
1235: struct nameidata nd;
1236:
1.268 elad 1237: if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT,
1238: KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0)
1.31 cgd 1239: return (error);
1.310 dsl 1240: NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
1.334 pooka 1241: SCARG(uap, path));
1.234 christos 1242: if ((error = change_dir(&nd, l)) != 0)
1.31 cgd 1243: return (error);
1.328 ad 1244:
1245: cwdi = p->p_cwdi;
1246: rw_enter(&cwdi->cwdi_lock, RW_WRITER);
1.134 thorpej 1247: if (cwdi->cwdi_rdir != NULL)
1248: vrele(cwdi->cwdi_rdir);
1.131 sommerfe 1249: vp = nd.ni_vp;
1.134 thorpej 1250: cwdi->cwdi_rdir = vp;
1.131 sommerfe 1251:
1252: /*
1253: * Prevent escaping from chroot by putting the root under
1254: * the working directory. Silently chdir to / if we aren't
1255: * already there.
1256: */
1.234 christos 1257: if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) {
1.131 sommerfe 1258: /*
1259: * XXX would be more failsafe to change directory to a
1260: * deadfs node here instead
1261: */
1.134 thorpej 1262: vrele(cwdi->cwdi_cdir);
1.131 sommerfe 1263: VREF(vp);
1.134 thorpej 1264: cwdi->cwdi_cdir = vp;
1.131 sommerfe 1265: }
1.328 ad 1266: rw_exit(&cwdi->cwdi_lock);
1.205 junyoung 1267:
1.31 cgd 1268: return (0);
1269: }
1270:
1271: /*
1272: * Common routine for chroot and chdir.
1273: */
1274: static int
1.234 christos 1275: change_dir(struct nameidata *ndp, struct lwp *l)
1.31 cgd 1276: {
1277: struct vnode *vp;
1278: int error;
1279:
1.63 christos 1280: if ((error = namei(ndp)) != 0)
1.31 cgd 1281: return (error);
1282: vp = ndp->ni_vp;
1283: if (vp->v_type != VDIR)
1284: error = ENOTDIR;
1285: else
1.332 pooka 1286: error = VOP_ACCESS(vp, VEXEC, l->l_cred);
1.205 junyoung 1287:
1.31 cgd 1288: if (error)
1.113 fvdl 1289: vput(vp);
1290: else
1291: VOP_UNLOCK(vp, 0);
1.31 cgd 1292: return (error);
1293: }
1294:
1295: /*
1296: * Check permissions, allocate an open file structure,
1297: * and call the device open routine if any.
1298: */
1.63 christos 1299: int
1.335 dsl 1300: sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval)
1.56 thorpej 1301: {
1.335 dsl 1302: /* {
1.74 cgd 1303: syscallarg(const char *) path;
1.35 cgd 1304: syscallarg(int) flags;
1305: syscallarg(int) mode;
1.335 dsl 1306: } */
1.179 thorpej 1307: struct proc *p = l->l_proc;
1.134 thorpej 1308: struct cwdinfo *cwdi = p->p_cwdi;
1.346 ad 1309: file_t *fp;
1.135 thorpej 1310: struct vnode *vp;
1.31 cgd 1311: int flags, cmode;
1312: int type, indx, error;
1313: struct flock lf;
1314: struct nameidata nd;
1315:
1.104 mycroft 1316: flags = FFLAGS(SCARG(uap, flags));
1317: if ((flags & (FREAD | FWRITE)) == 0)
1318: return (EINVAL);
1.346 ad 1319: if ((error = fd_allocfile(&fp, &indx)) != 0)
1.31 cgd 1320: return (error);
1.328 ad 1321: /* We're going to read cwdi->cwdi_cmask unlocked here. */
1.134 thorpej 1322: cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT;
1.331 pooka 1323: NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
1.334 pooka 1324: SCARG(uap, path));
1.194 jdolecek 1325: l->l_dupfd = -indx - 1; /* XXX check for fdopen */
1.63 christos 1326: if ((error = vn_open(&nd, flags, cmode)) != 0) {
1.346 ad 1327: fd_abort(p, fp, indx);
1.213 christos 1328: if ((error == EDUPFD || error == EMOVEFD) &&
1.194 jdolecek 1329: l->l_dupfd >= 0 && /* XXX from fdopen */
1.45 mycroft 1330: (error =
1.346 ad 1331: fd_dupopen(l->l_dupfd, &indx, flags, error)) == 0) {
1.45 mycroft 1332: *retval = indx;
1333: return (0);
1334: }
1335: if (error == ERESTART)
1.44 mycroft 1336: error = EINTR;
1.31 cgd 1337: return (error);
1338: }
1.331 pooka 1339:
1.194 jdolecek 1340: l->l_dupfd = 0;
1.31 cgd 1341: vp = nd.ni_vp;
1342: fp->f_flag = flags & FMASK;
1343: fp->f_type = DTYPE_VNODE;
1344: fp->f_ops = &vnops;
1.184 dsl 1345: fp->f_data = vp;
1.31 cgd 1346: if (flags & (O_EXLOCK | O_SHLOCK)) {
1347: lf.l_whence = SEEK_SET;
1348: lf.l_start = 0;
1349: lf.l_len = 0;
1350: if (flags & O_EXLOCK)
1351: lf.l_type = F_WRLCK;
1352: else
1353: lf.l_type = F_RDLCK;
1354: type = F_FLOCK;
1355: if ((flags & FNONBLOCK) == 0)
1356: type |= F_WAIT;
1.113 fvdl 1357: VOP_UNLOCK(vp, 0);
1.184 dsl 1358: error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type);
1.63 christos 1359: if (error) {
1.346 ad 1360: (void) vn_close(vp, fp->f_flag, fp->f_cred);
1361: fd_abort(p, fp, indx);
1.31 cgd 1362: return (error);
1363: }
1.113 fvdl 1364: vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1.346 ad 1365: atomic_or_uint(&fp->f_flag, FHASLOCK);
1.31 cgd 1366: }
1.113 fvdl 1367: VOP_UNLOCK(vp, 0);
1.31 cgd 1368: *retval = indx;
1.346 ad 1369: fd_affix(p, fp, indx);
1.31 cgd 1370: return (0);
1.137 wrstuden 1371: }
1372:
1.249 yamt 1373: static void
1374: vfs__fhfree(fhandle_t *fhp)
1375: {
1376: size_t fhsize;
1377:
1378: if (fhp == NULL) {
1379: return;
1380: }
1381: fhsize = FHANDLE_SIZE(fhp);
1382: kmem_free(fhp, fhsize);
1383: }
1384:
1.137 wrstuden 1385: /*
1.243 yamt 1386: * vfs_composefh: compose a filehandle.
1387: */
1388:
1389: int
1.244 martin 1390: vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size)
1.243 yamt 1391: {
1392: struct mount *mp;
1.250 yamt 1393: struct fid *fidp;
1.243 yamt 1394: int error;
1.250 yamt 1395: size_t needfhsize;
1396: size_t fidsize;
1.243 yamt 1397:
1398: mp = vp->v_mount;
1.250 yamt 1399: fidp = NULL;
1.252 martin 1400: if (*fh_size < FHANDLE_SIZE_MIN) {
1.250 yamt 1401: fidsize = 0;
1.244 martin 1402: } else {
1.250 yamt 1403: fidsize = *fh_size - offsetof(fhandle_t, fh_fid);
1404: if (fhp != NULL) {
1405: memset(fhp, 0, *fh_size);
1406: fhp->fh_fsid = mp->mnt_stat.f_fsidx;
1407: fidp = &fhp->fh_fid;
1408: }
1.244 martin 1409: }
1.250 yamt 1410: error = VFS_VPTOFH(vp, fidp, &fidsize);
1411: needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize);
1412: if (error == 0 && *fh_size < needfhsize) {
1413: error = E2BIG;
1414: }
1415: *fh_size = needfhsize;
1.243 yamt 1416: return error;
1417: }
1418:
1.249 yamt 1419: int
1420: vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp)
1421: {
1422: struct mount *mp;
1423: fhandle_t *fhp;
1424: size_t fhsize;
1425: size_t fidsize;
1426: int error;
1427:
1428: *fhpp = NULL;
1429: mp = vp->v_mount;
1.255 christos 1430: fidsize = 0;
1.249 yamt 1431: error = VFS_VPTOFH(vp, NULL, &fidsize);
1432: KASSERT(error != 0);
1433: if (error != E2BIG) {
1434: goto out;
1435: }
1.250 yamt 1436: fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize);
1.249 yamt 1437: fhp = kmem_zalloc(fhsize, KM_SLEEP);
1438: if (fhp == NULL) {
1439: error = ENOMEM;
1440: goto out;
1441: }
1442: fhp->fh_fsid = mp->mnt_stat.f_fsidx;
1443: error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize);
1444: if (error == 0) {
1445: KASSERT((FHANDLE_SIZE(fhp) == fhsize &&
1446: FHANDLE_FILEID(fhp)->fid_len == fidsize));
1447: *fhpp = fhp;
1448: } else {
1449: kmem_free(fhp, fhsize);
1450: }
1451: out:
1452: return error;
1453: }
1454:
1455: void
1456: vfs_composefh_free(fhandle_t *fhp)
1457: {
1458:
1459: vfs__fhfree(fhp);
1460: }
1461:
1.243 yamt 1462: /*
1.248 yamt 1463: * vfs_fhtovp: lookup a vnode by a filehandle.
1464: */
1465:
1466: int
1467: vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp)
1468: {
1469: struct mount *mp;
1470: int error;
1471:
1472: *vpp = NULL;
1473: mp = vfs_getvfs(FHANDLE_FSID(fhp));
1474: if (mp == NULL) {
1475: error = ESTALE;
1476: goto out;
1477: }
1478: if (mp->mnt_op->vfs_fhtovp == NULL) {
1479: error = EOPNOTSUPP;
1480: goto out;
1481: }
1482: error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp);
1483: out:
1484: return error;
1485: }
1486:
1487: /*
1.265 yamt 1488: * vfs_copyinfh_alloc: allocate and copyin a filehandle, given
1.263 martin 1489: * the needed size.
1490: */
1491:
1492: int
1.265 yamt 1493: vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp)
1.263 martin 1494: {
1495: fhandle_t *fhp;
1496: int error;
1497:
1498: *fhpp = NULL;
1.250 yamt 1499: if (fhsize > FHANDLE_SIZE_MAX) {
1500: return EINVAL;
1501: }
1.265 yamt 1502: if (fhsize < FHANDLE_SIZE_MIN) {
1503: return EINVAL;
1504: }
1.267 yamt 1505: again:
1.248 yamt 1506: fhp = kmem_alloc(fhsize, KM_SLEEP);
1507: if (fhp == NULL) {
1508: return ENOMEM;
1509: }
1510: error = copyin(ufhp, fhp, fhsize);
1511: if (error == 0) {
1.265 yamt 1512: /* XXX this check shouldn't be here */
1513: if (FHANDLE_SIZE(fhp) == fhsize) {
1.263 martin 1514: *fhpp = fhp;
1515: return 0;
1.267 yamt 1516: } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) {
1517: /*
1518: * a kludge for nfsv2 padded handles.
1519: */
1520: size_t sz;
1521:
1522: sz = FHANDLE_SIZE(fhp);
1523: kmem_free(fhp, fhsize);
1524: fhsize = sz;
1525: goto again;
1.264 yamt 1526: } else {
1.265 yamt 1527: /*
1528: * userland told us wrong size.
1529: */
1.263 martin 1530: error = EINVAL;
1.264 yamt 1531: }
1.248 yamt 1532: }
1.263 martin 1533: kmem_free(fhp, fhsize);
1.248 yamt 1534: return error;
1535: }
1536:
1537: void
1538: vfs_copyinfh_free(fhandle_t *fhp)
1539: {
1540:
1.249 yamt 1541: vfs__fhfree(fhp);
1.248 yamt 1542: }
1543:
1544: /*
1.137 wrstuden 1545: * Get file handle system call
1546: */
1547: int
1.335 dsl 1548: sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval)
1.137 wrstuden 1549: {
1.335 dsl 1550: /* {
1.137 wrstuden 1551: syscallarg(char *) fname;
1552: syscallarg(fhandle_t *) fhp;
1.244 martin 1553: syscallarg(size_t *) fh_size;
1.335 dsl 1554: } */
1.155 augustss 1555: struct vnode *vp;
1.244 martin 1556: fhandle_t *fh;
1.137 wrstuden 1557: int error;
1558: struct nameidata nd;
1.244 martin 1559: size_t sz;
1.249 yamt 1560: size_t usz;
1.137 wrstuden 1561:
1562: /*
1563: * Must be super user
1564: */
1.268 elad 1565: error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1566: 0, NULL, NULL, NULL);
1.137 wrstuden 1567: if (error)
1568: return (error);
1.310 dsl 1569: NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
1.334 pooka 1570: SCARG(uap, fname));
1.137 wrstuden 1571: error = namei(&nd);
1572: if (error)
1573: return (error);
1574: vp = nd.ni_vp;
1.249 yamt 1575: error = vfs_composefh_alloc(vp, &fh);
1.247 yamt 1576: vput(vp);
1.249 yamt 1577: if (error != 0) {
1578: goto out;
1579: }
1580: error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t));
1581: if (error != 0) {
1582: goto out;
1583: }
1584: sz = FHANDLE_SIZE(fh);
1585: error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t));
1586: if (error != 0) {
1587: goto out;
1.244 martin 1588: }
1.249 yamt 1589: if (usz >= sz) {
1590: error = copyout(fh, SCARG(uap, fhp), sz);
1591: } else {
1592: error = E2BIG;
1.244 martin 1593: }
1.249 yamt 1594: out:
1595: vfs_composefh_free(fh);
1.137 wrstuden 1596: return (error);
1597: }
1598:
1599: /*
1600: * Open a file given a file handle.
1601: *
1602: * Check permissions, allocate an open file structure,
1603: * and call the device open routine if any.
1604: */
1.265 yamt 1605:
1.137 wrstuden 1606: int
1.265 yamt 1607: dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags,
1608: register_t *retval)
1.137 wrstuden 1609: {
1.346 ad 1610: file_t *fp;
1.139 wrstuden 1611: struct vnode *vp = NULL;
1.257 ad 1612: kauth_cred_t cred = l->l_cred;
1.346 ad 1613: file_t *nfp;
1.137 wrstuden 1614: int type, indx, error=0;
1615: struct flock lf;
1616: struct vattr va;
1.248 yamt 1617: fhandle_t *fh;
1.265 yamt 1618: int flags;
1.346 ad 1619: proc_t *p;
1620:
1621: p = curproc;
1.137 wrstuden 1622:
1623: /*
1624: * Must be super user
1625: */
1.269 elad 1626: if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1.270 elad 1627: 0, NULL, NULL, NULL)))
1.137 wrstuden 1628: return (error);
1629:
1.265 yamt 1630: flags = FFLAGS(oflags);
1.137 wrstuden 1631: if ((flags & (FREAD | FWRITE)) == 0)
1632: return (EINVAL);
1633: if ((flags & O_CREAT))
1634: return (EINVAL);
1.346 ad 1635: if ((error = fd_allocfile(&nfp, &indx)) != 0)
1.137 wrstuden 1636: return (error);
1637: fp = nfp;
1.265 yamt 1638: error = vfs_copyinfh_alloc(ufhp, fhsize, &fh);
1.248 yamt 1639: if (error != 0) {
1.139 wrstuden 1640: goto bad;
1641: }
1.248 yamt 1642: error = vfs_fhtovp(fh, &vp);
1643: if (error != 0) {
1.139 wrstuden 1644: goto bad;
1645: }
1.137 wrstuden 1646:
1647: /* Now do an effective vn_open */
1648:
1649: if (vp->v_type == VSOCK) {
1650: error = EOPNOTSUPP;
1651: goto bad;
1652: }
1.333 yamt 1653: error = vn_openchk(vp, cred, flags);
1654: if (error != 0)
1655: goto bad;
1.137 wrstuden 1656: if (flags & O_TRUNC) {
1657: VOP_UNLOCK(vp, 0); /* XXX */
1658: vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */
1659: VATTR_NULL(&va);
1660: va.va_size = 0;
1.332 pooka 1661: error = VOP_SETATTR(vp, &va, cred);
1.197 hannken 1662: if (error)
1.137 wrstuden 1663: goto bad;
1664: }
1.332 pooka 1665: if ((error = VOP_OPEN(vp, flags, cred)) != 0)
1.137 wrstuden 1666: goto bad;
1.346 ad 1667: if (flags & FWRITE) {
1668: mutex_enter(&vp->v_interlock);
1.137 wrstuden 1669: vp->v_writecount++;
1.346 ad 1670: mutex_exit(&vp->v_interlock);
1671: }
1.137 wrstuden 1672:
1673: /* done with modified vn_open, now finish what sys_open does. */
1674:
1675: fp->f_flag = flags & FMASK;
1676: fp->f_type = DTYPE_VNODE;
1677: fp->f_ops = &vnops;
1.184 dsl 1678: fp->f_data = vp;
1.137 wrstuden 1679: if (flags & (O_EXLOCK | O_SHLOCK)) {
1680: lf.l_whence = SEEK_SET;
1681: lf.l_start = 0;
1682: lf.l_len = 0;
1683: if (flags & O_EXLOCK)
1684: lf.l_type = F_WRLCK;
1685: else
1686: lf.l_type = F_RDLCK;
1687: type = F_FLOCK;
1688: if ((flags & FNONBLOCK) == 0)
1689: type |= F_WAIT;
1690: VOP_UNLOCK(vp, 0);
1.184 dsl 1691: error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type);
1.137 wrstuden 1692: if (error) {
1.346 ad 1693: (void) vn_close(vp, fp->f_flag, fp->f_cred);
1694: fd_abort(p, fp, indx);
1.137 wrstuden 1695: return (error);
1696: }
1697: vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1.346 ad 1698: atomic_or_uint(&fp->f_flag, FHASLOCK);
1.137 wrstuden 1699: }
1700: VOP_UNLOCK(vp, 0);
1701: *retval = indx;
1.346 ad 1702: fd_affix(p, fp, indx);
1.248 yamt 1703: vfs_copyinfh_free(fh);
1.137 wrstuden 1704: return (0);
1.139 wrstuden 1705:
1.137 wrstuden 1706: bad:
1.346 ad 1707: fd_abort(p, fp, indx);
1.139 wrstuden 1708: if (vp != NULL)
1709: vput(vp);
1.248 yamt 1710: vfs_copyinfh_free(fh);
1.137 wrstuden 1711: return (error);
1712: }
1713:
1714: int
1.335 dsl 1715: sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval)
1.137 wrstuden 1716: {
1.335 dsl 1717: /* {
1.263 martin 1718: syscallarg(const void *) fhp;
1719: syscallarg(size_t) fh_size;
1.265 yamt 1720: syscallarg(int) flags;
1.335 dsl 1721: } */
1.265 yamt 1722:
1723: return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size),
1724: SCARG(uap, flags), retval);
1725: }
1726:
1.306 dsl 1727: int
1728: do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb)
1729: {
1.137 wrstuden 1730: int error;
1.248 yamt 1731: fhandle_t *fh;
1.137 wrstuden 1732: struct vnode *vp;
1733:
1734: /*
1735: * Must be super user
1736: */
1.268 elad 1737: if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1.270 elad 1738: 0, NULL, NULL, NULL)))
1.137 wrstuden 1739: return (error);
1740:
1.265 yamt 1741: error = vfs_copyinfh_alloc(ufhp, fhsize, &fh);
1.306 dsl 1742: if (error != 0)
1743: return error;
1744:
1.248 yamt 1745: error = vfs_fhtovp(fh, &vp);
1.306 dsl 1746: vfs_copyinfh_free(fh);
1747: if (error != 0)
1748: return error;
1749:
1.346 ad 1750: error = vn_stat(vp, sb);
1.137 wrstuden 1751: vput(vp);
1.248 yamt 1752: return error;
1.137 wrstuden 1753: }
1754:
1.265 yamt 1755:
1.137 wrstuden 1756: /* ARGSUSED */
1757: int
1.383 christos 1758: sys___fhstat50(struct lwp *l, const struct sys___fhstat50_args *uap, register_t *retval)
1.137 wrstuden 1759: {
1.335 dsl 1760: /* {
1.265 yamt 1761: syscallarg(const void *) fhp;
1.263 martin 1762: syscallarg(size_t) fh_size;
1.265 yamt 1763: syscallarg(struct stat *) sb;
1.335 dsl 1764: } */
1.306 dsl 1765: struct stat sb;
1766: int error;
1.265 yamt 1767:
1.306 dsl 1768: error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb);
1769: if (error)
1770: return error;
1771: return copyout(&sb, SCARG(uap, sb), sizeof(sb));
1.265 yamt 1772: }
1773:
1.306 dsl 1774: int
1775: do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb,
1776: int flags)
1777: {
1.248 yamt 1778: fhandle_t *fh;
1.137 wrstuden 1779: struct mount *mp;
1780: struct vnode *vp;
1781: int error;
1782:
1783: /*
1784: * Must be super user
1785: */
1.268 elad 1786: if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1.270 elad 1787: 0, NULL, NULL, NULL)))
1.206 christos 1788: return error;
1.137 wrstuden 1789:
1.265 yamt 1790: error = vfs_copyinfh_alloc(ufhp, fhsize, &fh);
1.306 dsl 1791: if (error != 0)
1792: return error;
1793:
1.248 yamt 1794: error = vfs_fhtovp(fh, &vp);
1.306 dsl 1795: vfs_copyinfh_free(fh);
1796: if (error != 0)
1797: return error;
1798:
1.137 wrstuden 1799: mp = vp->v_mount;
1.306 dsl 1800: error = dostatvfs(mp, sb, l, flags, 1);
1.137 wrstuden 1801: vput(vp);
1.241 yamt 1802: return error;
1.31 cgd 1803: }
1804:
1.265 yamt 1805: /* ARGSUSED */
1806: int
1.335 dsl 1807: sys___fhstatvfs140(struct lwp *l, const struct sys___fhstatvfs140_args *uap, register_t *retval)
1.265 yamt 1808: {
1.335 dsl 1809: /* {
1.266 yamt 1810: syscallarg(const void *) fhp;
1.265 yamt 1811: syscallarg(size_t) fh_size;
1812: syscallarg(struct statvfs *) buf;
1813: syscallarg(int) flags;
1.335 dsl 1814: } */
1.306 dsl 1815: struct statvfs *sb = STATVFSBUF_GET();
1816: int error;
1.265 yamt 1817:
1.306 dsl 1818: error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb,
1819: SCARG(uap, flags));
1820: if (error == 0)
1821: error = copyout(sb, SCARG(uap, buf), sizeof(*sb));
1822: STATVFSBUF_PUT(sb);
1823: return error;
1.265 yamt 1824: }
1825:
1.31 cgd 1826: /*
1827: * Create a special file.
1828: */
1829: /* ARGSUSED */
1.63 christos 1830: int
1.383 christos 1831: sys___mknod50(struct lwp *l, const struct sys___mknod50_args *uap,
1832: register_t *retval)
1.56 thorpej 1833: {
1.335 dsl 1834: /* {
1.74 cgd 1835: syscallarg(const char *) path;
1.383 christos 1836: syscallarg(mode_t) mode;
1837: syscallarg(dev_t) dev;
1.335 dsl 1838: } */
1.383 christos 1839: return do_sys_mknod(l, SCARG(uap, path), SCARG(uap, mode),
1840: SCARG(uap, dev), retval);
1841: }
1842:
1843: int
1844: do_sys_mknod(struct lwp *l, const char *pathname, mode_t mode, dev_t dev,
1845: register_t *retval)
1846: {
1.179 thorpej 1847: struct proc *p = l->l_proc;
1.155 augustss 1848: struct vnode *vp;
1.31 cgd 1849: struct vattr vattr;
1.301 pooka 1850: int error, optype;
1.315 christos 1851: struct nameidata nd;
1.314 christos 1852: char *path;
1.315 christos 1853: const char *cpath;
1854: enum uio_seg seg = UIO_USERSPACE;
1.31 cgd 1855:
1.268 elad 1856: if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD,
1857: 0, NULL, NULL, NULL)) != 0)
1.31 cgd 1858: return (error);
1.301 pooka 1859:
1860: optype = VOP_MKNOD_DESCOFFSET;
1.315 christos 1861:
1.383 christos 1862: VERIEXEC_PATH_GET(pathname, seg, cpath, path);
1.334 pooka 1863: NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, seg, cpath);
1.315 christos 1864:
1.63 christos 1865: if ((error = namei(&nd)) != 0)
1.314 christos 1866: goto out;
1.31 cgd 1867: vp = nd.ni_vp;
1.43 mycroft 1868: if (vp != NULL)
1.31 cgd 1869: error = EEXIST;
1.43 mycroft 1870: else {
1871: VATTR_NULL(&vattr);
1.328 ad 1872: /* We will read cwdi->cwdi_cmask unlocked. */
1.383 christos 1873: vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1874: vattr.va_rdev = dev;
1.43 mycroft 1875:
1.383 christos 1876: switch (mode & S_IFMT) {
1.43 mycroft 1877: case S_IFMT: /* used by badsect to flag bad sectors */
1878: vattr.va_type = VBAD;
1879: break;
1880: case S_IFCHR:
1881: vattr.va_type = VCHR;
1882: break;
1883: case S_IFBLK:
1884: vattr.va_type = VBLK;
1885: break;
1886: case S_IFWHT:
1.301 pooka 1887: optype = VOP_WHITEOUT_DESCOFFSET;
1888: break;
1889: case S_IFREG:
1.314 christos 1890: #if NVERIEXEC > 0
1891: error = veriexec_openchk(l, nd.ni_vp, nd.ni_dirp,
1892: O_CREAT);
1893: #endif /* NVERIEXEC > 0 */
1.301 pooka 1894: vattr.va_type = VREG;
1.302 pooka 1895: vattr.va_rdev = VNOVAL;
1.301 pooka 1896: optype = VOP_CREATE_DESCOFFSET;
1.43 mycroft 1897: break;
1898: default:
1899: error = EINVAL;
1900: break;
1901: }
1.31 cgd 1902: }
1.43 mycroft 1903: if (!error) {
1.301 pooka 1904: switch (optype) {
1905: case VOP_WHITEOUT_DESCOFFSET:
1.43 mycroft 1906: error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1907: if (error)
1908: VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1909: vput(nd.ni_dvp);
1.301 pooka 1910: break;
1911:
1912: case VOP_MKNOD_DESCOFFSET:
1.43 mycroft 1913: error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1914: &nd.ni_cnd, &vattr);
1.168 assar 1915: if (error == 0)
1916: vput(nd.ni_vp);
1.301 pooka 1917: break;
1918:
1919: case VOP_CREATE_DESCOFFSET:
1920: error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp,
1921: &nd.ni_cnd, &vattr);
1922: if (error == 0)
1923: vput(nd.ni_vp);
1924: break;
1.43 mycroft 1925: }
1926: } else {
1927: VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1928: if (nd.ni_dvp == vp)
1929: vrele(nd.ni_dvp);
1930: else
1931: vput(nd.ni_dvp);
1932: if (vp)
1933: vrele(vp);
1.31 cgd 1934: }
1.314 christos 1935: out:
1.315 christos 1936: VERIEXEC_PATH_PUT(path);
1.31 cgd 1937: return (error);
1938: }
1939:
1940: /*
1941: * Create a named pipe.
1942: */
1943: /* ARGSUSED */
1.63 christos 1944: int
1.335 dsl 1945: sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval)
1.56 thorpej 1946: {
1.335 dsl 1947: /* {
1.74 cgd 1948: syscallarg(const char *) path;
1.35 cgd 1949: syscallarg(int) mode;
1.335 dsl 1950: } */
1.179 thorpej 1951: struct proc *p = l->l_proc;
1.31 cgd 1952: struct vattr vattr;
1953: int error;
1954: struct nameidata nd;
1955:
1.334 pooka 1956: NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE,
1957: SCARG(uap, path));
1.63 christos 1958: if ((error = namei(&nd)) != 0)
1.31 cgd 1959: return (error);
1960: if (nd.ni_vp != NULL) {
1961: VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1962: if (nd.ni_dvp == nd.ni_vp)
1963: vrele(nd.ni_dvp);
1964: else
1965: vput(nd.ni_dvp);
1966: vrele(nd.ni_vp);
1967: return (EEXIST);
1968: }
1969: VATTR_NULL(&vattr);
1970: vattr.va_type = VFIFO;
1.328 ad 1971: /* We will read cwdi->cwdi_cmask unlocked. */
1.134 thorpej 1972: vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1.168 assar 1973: error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1974: if (error == 0)
1975: vput(nd.ni_vp);
1976: return (error);
1.31 cgd 1977: }
1978:
1979: /*
1980: * Make a hard file link.
1981: */
1982: /* ARGSUSED */
1.63 christos 1983: int
1.335 dsl 1984: sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval)
1.56 thorpej 1985: {
1.335 dsl 1986: /* {
1.74 cgd 1987: syscallarg(const char *) path;
1988: syscallarg(const char *) link;
1.335 dsl 1989: } */
1.155 augustss 1990: struct vnode *vp;
1.31 cgd 1991: struct nameidata nd;
1992: int error;
1993:
1.331 pooka 1994: NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
1.334 pooka 1995: SCARG(uap, path));
1.63 christos 1996: if ((error = namei(&nd)) != 0)
1.31 cgd 1997: return (error);
1998: vp = nd.ni_vp;
1.331 pooka 1999: NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE,
1.334 pooka 2000: SCARG(uap, link));
1.66 mycroft 2001: if ((error = namei(&nd)) != 0)
2002: goto out;
2003: if (nd.ni_vp) {
2004: VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2005: if (nd.ni_dvp == nd.ni_vp)
2006: vrele(nd.ni_dvp);
2007: else
2008: vput(nd.ni_dvp);
2009: vrele(nd.ni_vp);
2010: error = EEXIST;
2011: goto out;
1.31 cgd 2012: }
1.66 mycroft 2013: error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
2014: out:
1.31 cgd 2015: vrele(vp);
2016: return (error);
2017: }
2018:
2019: /*
2020: * Make a symbolic link.
2021: */
2022: /* ARGSUSED */
1.63 christos 2023: int
1.335 dsl 2024: sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval)
1.56 thorpej 2025: {
1.335 dsl 2026: /* {
1.74 cgd 2027: syscallarg(const char *) path;
2028: syscallarg(const char *) link;
1.335 dsl 2029: } */
1.179 thorpej 2030: struct proc *p = l->l_proc;
1.31 cgd 2031: struct vattr vattr;
2032: char *path;
2033: int error;
2034: struct nameidata nd;
2035:
1.161 thorpej 2036: path = PNBUF_GET();
1.63 christos 2037: error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL);
2038: if (error)
1.43 mycroft 2039: goto out;
1.331 pooka 2040: NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE,
1.334 pooka 2041: SCARG(uap, link));
1.63 christos 2042: if ((error = namei(&nd)) != 0)
1.43 mycroft 2043: goto out;
1.31 cgd 2044: if (nd.ni_vp) {
2045: VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2046: if (nd.ni_dvp == nd.ni_vp)
2047: vrele(nd.ni_dvp);
2048: else
2049: vput(nd.ni_dvp);
2050: vrele(nd.ni_vp);
2051: error = EEXIST;
1.43 mycroft 2052: goto out;
1.31 cgd 2053: }
2054: VATTR_NULL(&vattr);
1.230 jmmv 2055: vattr.va_type = VLNK;
1.328 ad 2056: /* We will read cwdi->cwdi_cmask unlocked. */
1.134 thorpej 2057: vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask;
1.31 cgd 2058: error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path);
1.168 assar 2059: if (error == 0)
2060: vput(nd.ni_vp);
1.43 mycroft 2061: out:
1.161 thorpej 2062: PNBUF_PUT(path);
1.31 cgd 2063: return (error);
2064: }
2065:
2066: /*
1.43 mycroft 2067: * Delete a whiteout from the filesystem.
2068: */
2069: /* ARGSUSED */
1.63 christos 2070: int
1.335 dsl 2071: sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval)
1.56 thorpej 2072: {
1.335 dsl 2073: /* {
1.74 cgd 2074: syscallarg(const char *) path;
1.335 dsl 2075: } */
1.43 mycroft 2076: int error;
2077: struct nameidata nd;
2078:
1.331 pooka 2079: NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT,
1.334 pooka 2080: UIO_USERSPACE, SCARG(uap, path));
1.43 mycroft 2081: error = namei(&nd);
2082: if (error)
2083: return (error);
2084:
2085: if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
2086: VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2087: if (nd.ni_dvp == nd.ni_vp)
2088: vrele(nd.ni_dvp);
2089: else
2090: vput(nd.ni_dvp);
2091: if (nd.ni_vp)
2092: vrele(nd.ni_vp);
2093: return (EEXIST);
2094: }
1.63 christos 2095: if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0)
1.43 mycroft 2096: VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2097: vput(nd.ni_dvp);
2098: return (error);
2099: }
2100:
2101: /*
1.31 cgd 2102: * Delete a name from the filesystem.
2103: */
2104: /* ARGSUSED */
1.63 christos 2105: int
1.335 dsl 2106: sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval)
1.56 thorpej 2107: {
1.335 dsl 2108: /* {
1.74 cgd 2109: syscallarg(const char *) path;
1.335 dsl 2110: } */
1.336 ad 2111:
2112: return do_sys_unlink(SCARG(uap, path), UIO_USERSPACE);
2113: }
2114:
2115: int
2116: do_sys_unlink(const char *arg, enum uio_seg seg)
2117: {
1.155 augustss 2118: struct vnode *vp;
1.31 cgd 2119: int error;
2120: struct nameidata nd;
1.336 ad 2121: kauth_cred_t cred;
1.315 christos 2122: char *path;
2123: const char *cpath;
1.31 cgd 2124:
1.336 ad 2125: VERIEXEC_PATH_GET(arg, seg, cpath, path);
1.334 pooka 2126: NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, seg, cpath);
1.313 elad 2127:
1.63 christos 2128: if ((error = namei(&nd)) != 0)
1.313 elad 2129: goto out;
1.31 cgd 2130: vp = nd.ni_vp;
2131:
1.66 mycroft 2132: /*
2133: * The root of a mounted filesystem cannot be deleted.
2134: */
1.329 ad 2135: if (vp->v_vflag & VV_ROOT) {
1.43 mycroft 2136: VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2137: if (nd.ni_dvp == vp)
2138: vrele(nd.ni_dvp);
2139: else
2140: vput(nd.ni_dvp);
1.66 mycroft 2141: vput(vp);
2142: error = EBUSY;
2143: goto out;
1.31 cgd 2144: }
1.219 blymn 2145:
1.256 elad 2146: #if NVERIEXEC > 0
1.222 elad 2147: /* Handle remove requests for veriexec entries. */
1.336 ad 2148: if ((error = veriexec_removechk(curlwp, nd.ni_vp, nd.ni_dirp)) != 0) {
1.219 blymn 2149: VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2150: if (nd.ni_dvp == vp)
2151: vrele(nd.ni_dvp);
2152: else
2153: vput(nd.ni_dvp);
2154: vput(vp);
2155: goto out;
2156: }
1.256 elad 2157: #endif /* NVERIEXEC > 0 */
2158:
1.336 ad 2159: cred = kauth_cred_get();
1.253 elad 2160: #ifdef FILEASSOC
1.278 elad 2161: (void)fileassoc_file_delete(vp);
1.253 elad 2162: #endif /* FILEASSOC */
1.66 mycroft 2163: error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
2164: out:
1.315 christos 2165: VERIEXEC_PATH_PUT(path);
1.31 cgd 2166: return (error);
2167: }
2168:
2169: /*
2170: * Reposition read/write file offset.
2171: */
1.63 christos 2172: int
1.335 dsl 2173: sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval)
1.56 thorpej 2174: {
1.335 dsl 2175: /* {
1.35 cgd 2176: syscallarg(int) fd;
2177: syscallarg(int) pad;
2178: syscallarg(off_t) offset;
2179: syscallarg(int) whence;
1.335 dsl 2180: } */
1.257 ad 2181: kauth_cred_t cred = l->l_cred;
1.346 ad 2182: file_t *fp;
1.84 kleink 2183: struct vnode *vp;
1.31 cgd 2184: struct vattr vattr;
1.155 augustss 2185: off_t newoff;
1.346 ad 2186: int error, fd;
2187:
2188: fd = SCARG(uap, fd);
1.31 cgd 2189:
1.346 ad 2190: if ((fp = fd_getfile(fd)) == NULL)
1.31 cgd 2191: return (EBADF);
1.84 kleink 2192:
1.346 ad 2193: vp = fp->f_data;
1.135 thorpej 2194: if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2195: error = ESPIPE;
2196: goto out;
2197: }
1.84 kleink 2198:
1.35 cgd 2199: switch (SCARG(uap, whence)) {
1.92 kleink 2200: case SEEK_CUR:
1.84 kleink 2201: newoff = fp->f_offset + SCARG(uap, offset);
1.31 cgd 2202: break;
1.92 kleink 2203: case SEEK_END:
1.332 pooka 2204: error = VOP_GETATTR(vp, &vattr, cred);
1.328 ad 2205: if (error) {
1.135 thorpej 2206: goto out;
1.328 ad 2207: }
1.84 kleink 2208: newoff = SCARG(uap, offset) + vattr.va_size;
1.31 cgd 2209: break;
1.92 kleink 2210: case SEEK_SET:
1.84 kleink 2211: newoff = SCARG(uap, offset);
1.31 cgd 2212: break;
2213: default:
1.135 thorpej 2214: error = EINVAL;
2215: goto out;
1.31 cgd 2216: }
1.328 ad 2217: if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) {
2218: *(off_t *)retval = fp->f_offset = newoff;
2219: }
1.135 thorpej 2220: out:
1.346 ad 2221: fd_putfile(fd);
1.135 thorpej 2222: return (error);
1.120 thorpej 2223: }
2224:
2225: /*
2226: * Positional read system call.
2227: */
2228: int
1.335 dsl 2229: sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval)
1.120 thorpej 2230: {
1.335 dsl 2231: /* {
1.120 thorpej 2232: syscallarg(int) fd;
2233: syscallarg(void *) buf;
2234: syscallarg(size_t) nbyte;
2235: syscallarg(off_t) offset;
1.335 dsl 2236: } */
1.346 ad 2237: file_t *fp;
1.120 thorpej 2238: struct vnode *vp;
2239: off_t offset;
2240: int error, fd = SCARG(uap, fd);
2241:
1.346 ad 2242: if ((fp = fd_getfile(fd)) == NULL)
1.166 thorpej 2243: return (EBADF);
2244:
1.183 pk 2245: if ((fp->f_flag & FREAD) == 0) {
1.346 ad 2246: fd_putfile(fd);
1.120 thorpej 2247: return (EBADF);
1.183 pk 2248: }
1.120 thorpej 2249:
1.346 ad 2250: vp = fp->f_data;
1.135 thorpej 2251: if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2252: error = ESPIPE;
2253: goto out;
2254: }
1.120 thorpej 2255:
2256: offset = SCARG(uap, offset);
2257:
2258: /*
2259: * XXX This works because no file systems actually
2260: * XXX take any action on the seek operation.
2261: */
2262: if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1.135 thorpej 2263: goto out;
1.120 thorpej 2264:
1.135 thorpej 2265: /* dofileread() will unuse the descriptor for us */
1.328 ad 2266: return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
1.120 thorpej 2267: &offset, 0, retval));
1.135 thorpej 2268:
2269: out:
1.346 ad 2270: fd_putfile(fd);
1.135 thorpej 2271: return (error);
1.120 thorpej 2272: }
2273:
2274: /*
2275: * Positional scatter read system call.
2276: */
2277: int
1.335 dsl 2278: sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval)
1.120 thorpej 2279: {
1.335 dsl 2280: /* {
1.120 thorpej 2281: syscallarg(int) fd;
2282: syscallarg(const struct iovec *) iovp;
2283: syscallarg(int) iovcnt;
2284: syscallarg(off_t) offset;
1.335 dsl 2285: } */
2286: off_t offset = SCARG(uap, offset);
1.120 thorpej 2287:
1.328 ad 2288: return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp),
1.335 dsl 2289: SCARG(uap, iovcnt), &offset, 0, retval);
1.120 thorpej 2290: }
2291:
2292: /*
2293: * Positional write system call.
2294: */
2295: int
1.335 dsl 2296: sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval)
1.120 thorpej 2297: {
1.335 dsl 2298: /* {
1.120 thorpej 2299: syscallarg(int) fd;
2300: syscallarg(const void *) buf;
2301: syscallarg(size_t) nbyte;
2302: syscallarg(off_t) offset;
1.335 dsl 2303: } */
1.346 ad 2304: file_t *fp;
1.120 thorpej 2305: struct vnode *vp;
2306: off_t offset;
2307: int error, fd = SCARG(uap, fd);
2308:
1.346 ad 2309: if ((fp = fd_getfile(fd)) == NULL)
1.166 thorpej 2310: return (EBADF);
2311:
1.183 pk 2312: if ((fp->f_flag & FWRITE) == 0) {
1.346 ad 2313: fd_putfile(fd);
1.120 thorpej 2314: return (EBADF);
1.183 pk 2315: }
1.120 thorpej 2316:
1.346 ad 2317: vp = fp->f_data;
1.135 thorpej 2318: if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2319: error = ESPIPE;
2320: goto out;
2321: }
1.120 thorpej 2322:
2323: offset = SCARG(uap, offset);
2324:
2325: /*
2326: * XXX This works because no file systems actually
2327: * XXX take any action on the seek operation.
2328: */
2329: if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1.135 thorpej 2330: goto out;
1.120 thorpej 2331:
1.135 thorpej 2332: /* dofilewrite() will unuse the descriptor for us */
1.328 ad 2333: return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
1.120 thorpej 2334: &offset, 0, retval));
1.135 thorpej 2335:
2336: out:
1.346 ad 2337: fd_putfile(fd);
1.135 thorpej 2338: return (error);
1.120 thorpej 2339: }
2340:
2341: /*
2342: * Positional gather write system call.
2343: */
2344: int
1.335 dsl 2345: sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval)
1.120 thorpej 2346: {
1.335 dsl 2347: /* {
1.120 thorpej 2348: syscallarg(int) fd;
2349: syscallarg(const struct iovec *) iovp;
2350: syscallarg(int) iovcnt;
2351: syscallarg(off_t) offset;
1.335 dsl 2352: } */
2353: off_t offset = SCARG(uap, offset);
1.120 thorpej 2354:
1.328 ad 2355: return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp),
1.335 dsl 2356: SCARG(uap, iovcnt), &offset, 0, retval);
1.31 cgd 2357: }
2358:
2359: /*
2360: * Check access permissions.
2361: */
1.63 christos 2362: int
1.335 dsl 2363: sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval)
1.56 thorpej 2364: {
1.335 dsl 2365: /* {
1.74 cgd 2366: syscallarg(const char *) path;
1.35 cgd 2367: syscallarg(int) flags;
1.335 dsl 2368: } */
1.242 elad 2369: kauth_cred_t cred;
1.155 augustss 2370: struct vnode *vp;
1.169 christos 2371: int error, flags;
1.31 cgd 2372: struct nameidata nd;
2373:
1.257 ad 2374: cred = kauth_cred_dup(l->l_cred);
2375: kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred));
2376: kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred));
1.310 dsl 2377: NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
1.334 pooka 2378: SCARG(uap, path));
1.169 christos 2379: /* Override default credentials */
2380: nd.ni_cnd.cn_cred = cred;
1.63 christos 2381: if ((error = namei(&nd)) != 0)
1.169 christos 2382: goto out;
1.31 cgd 2383: vp = nd.ni_vp;
2384:
2385: /* Flags == 0 means only check for existence. */
1.35 cgd 2386: if (SCARG(uap, flags)) {
1.31 cgd 2387: flags = 0;
1.35 cgd 2388: if (SCARG(uap, flags) & R_OK)
1.31 cgd 2389: flags |= VREAD;
1.35 cgd 2390: if (SCARG(uap, flags) & W_OK)
1.31 cgd 2391: flags |= VWRITE;
1.89 mycroft 2392: if (SCARG(uap, flags) & X_OK)
2393: flags |= VEXEC;
1.138 is 2394:
1.332 pooka 2395: error = VOP_ACCESS(vp, flags, cred);
1.138 is 2396: if (!error && (flags & VWRITE))
2397: error = vn_writechk(vp);
1.31 cgd 2398: }
2399: vput(vp);
1.169 christos 2400: out:
1.242 elad 2401: kauth_cred_free(cred);
1.31 cgd 2402: return (error);
2403: }
2404:
2405: /*
1.306 dsl 2406: * Common code for all sys_stat functions, including compat versions.
2407: */
2408: int
1.346 ad 2409: do_sys_stat(const char *path, unsigned int nd_flags, struct stat *sb)
1.306 dsl 2410: {
2411: int error;
2412: struct nameidata nd;
2413:
1.331 pooka 2414: NDINIT(&nd, LOOKUP, nd_flags | LOCKLEAF | TRYEMULROOT,
1.334 pooka 2415: UIO_USERSPACE, path);
1.306 dsl 2416: error = namei(&nd);
2417: if (error != 0)
2418: return error;
1.346 ad 2419: error = vn_stat(nd.ni_vp, sb);
1.306 dsl 2420: vput(nd.ni_vp);
2421: return error;
2422: }
2423:
2424: /*
1.31 cgd 2425: * Get file status; this version follows links.
2426: */
2427: /* ARGSUSED */
1.63 christos 2428: int
1.383 christos 2429: sys___stat50(struct lwp *l, const struct sys___stat50_args *uap, register_t *retval)
1.56 thorpej 2430: {
1.335 dsl 2431: /* {
1.74 cgd 2432: syscallarg(const char *) path;
1.35 cgd 2433: syscallarg(struct stat *) ub;
1.335 dsl 2434: } */
1.31 cgd 2435: struct stat sb;
2436: int error;
2437:
1.346 ad 2438: error = do_sys_stat(SCARG(uap, path), FOLLOW, &sb);
1.31 cgd 2439: if (error)
1.306 dsl 2440: return error;
2441: return copyout(&sb, SCARG(uap, ub), sizeof(sb));
1.31 cgd 2442: }
2443:
2444: /*
2445: * Get file status; this version does not follow links.
2446: */
2447: /* ARGSUSED */
1.63 christos 2448: int
1.383 christos 2449: sys___lstat50(struct lwp *l, const struct sys___lstat50_args *uap, register_t *retval)
1.56 thorpej 2450: {
1.335 dsl 2451: /* {
1.74 cgd 2452: syscallarg(const char *) path;
1.35 cgd 2453: syscallarg(struct stat *) ub;
1.335 dsl 2454: } */
1.65 mycroft 2455: struct stat sb;
1.31 cgd 2456: int error;
2457:
1.346 ad 2458: error = do_sys_stat(SCARG(uap, path), NOFOLLOW, &sb);
1.64 jtc 2459: if (error)
1.306 dsl 2460: return error;
2461: return copyout(&sb, SCARG(uap, ub), sizeof(sb));
1.31 cgd 2462: }
2463:
2464: /*
2465: * Get configurable pathname variables.
2466: */
2467: /* ARGSUSED */
1.63 christos 2468: int
1.335 dsl 2469: sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval)
1.56 thorpej 2470: {
1.335 dsl 2471: /* {
1.74 cgd 2472: syscallarg(const char *) path;
1.35 cgd 2473: syscallarg(int) name;
1.335 dsl 2474: } */
1.31 cgd 2475: int error;
2476: struct nameidata nd;
2477:
1.310 dsl 2478: NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
1.334 pooka 2479: SCARG(uap, path));
1.63 christos 2480: if ((error = namei(&nd)) != 0)
1.31 cgd 2481: return (error);
1.35 cgd 2482: error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval);
1.31 cgd 2483: vput(nd.ni_vp);
2484: return (error);
2485: }
2486:
2487: /*
2488: * Return target name of a symbolic link.
2489: */
2490: /* ARGSUSED */
1.63 christos 2491: int
1.335 dsl 2492: sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, register_t *retval)
1.56 thorpej 2493: {
1.335 dsl 2494: /* {
1.74 cgd 2495: syscallarg(const char *) path;
1.35 cgd 2496: syscallarg(char *) buf;
1.115 kleink 2497: syscallarg(size_t) count;
1.335 dsl 2498: } */
1.155 augustss 2499: struct vnode *vp;
1.31 cgd 2500: struct iovec aiov;
2501: struct uio auio;
2502: int error;
2503: struct nameidata nd;
2504:
1.310 dsl 2505: NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
1.334 pooka 2506: SCARG(uap, path));
1.63 christos 2507: if ((error = namei(&nd)) != 0)
1.31 cgd 2508: return (error);
2509: vp = nd.ni_vp;
2510: if (vp->v_type != VLNK)
2511: error = EINVAL;
1.106 enami 2512: else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) ||
1.332 pooka 2513: (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) {
1.35 cgd 2514: aiov.iov_base = SCARG(uap, buf);
2515: aiov.iov_len = SCARG(uap, count);
1.31 cgd 2516: auio.uio_iov = &aiov;
2517: auio.uio_iovcnt = 1;
2518: auio.uio_offset = 0;
2519: auio.uio_rw = UIO_READ;
1.238 yamt 2520: KASSERT(l == curlwp);
2521: auio.uio_vmspace = l->l_proc->p_vmspace;
1.35 cgd 2522: auio.uio_resid = SCARG(uap, count);
1.257 ad 2523: error = VOP_READLINK(vp, &auio, l->l_cred);
1.31 cgd 2524: }
2525: vput(vp);
1.35 cgd 2526: *retval = SCARG(uap, count) - auio.uio_resid;
1.31 cgd 2527: return (error);
2528: }
2529:
2530: /*
2531: * Change flags of a file given a path name.
2532: */
2533: /* ARGSUSED */
1.63 christos 2534: int
1.335 dsl 2535: sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval)
1.56 thorpej 2536: {
1.335 dsl 2537: /* {
1.74 cgd 2538: syscallarg(const char *) path;
2539: syscallarg(u_long) flags;
1.335 dsl 2540: } */
1.155 augustss 2541: struct vnode *vp;
1.31 cgd 2542: int error;
2543: struct nameidata nd;
2544:
1.331 pooka 2545: NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
1.334 pooka 2546: SCARG(uap, path));
1.63 christos 2547: if ((error = namei(&nd)) != 0)
1.31 cgd 2548: return (error);
2549: vp = nd.ni_vp;
1.234 christos 2550: error = change_flags(vp, SCARG(uap, flags), l);
1.31 cgd 2551: vput(vp);
2552: return (error);
2553: }
2554:
2555: /*
2556: * Change flags of a file given a file descriptor.
2557: */
2558: /* ARGSUSED */
1.63 christos 2559: int
1.335 dsl 2560: sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval)
1.56 thorpej 2561: {
1.335 dsl 2562: /* {
1.35 cgd 2563: syscallarg(int) fd;
1.74 cgd 2564: syscallarg(u_long) flags;
1.335 dsl 2565: } */
1.31 cgd 2566: struct vnode *vp;
1.346 ad 2567: file_t *fp;
1.31 cgd 2568: int error;
2569:
1.346 ad 2570: /* fd_getvnode() will use the descriptor for us */
2571: if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
1.31 cgd 2572: return (error);
1.346 ad 2573: vp = fp->f_data;
1.234 christos 2574: error = change_flags(vp, SCARG(uap, flags), l);
1.113 fvdl 2575: VOP_UNLOCK(vp, 0);
1.346 ad 2576: fd_putfile(SCARG(uap, fd));
1.156 mrg 2577: return (error);
2578: }
2579:
2580: /*
1.163 enami 2581: * Change flags of a file given a path name; this version does
1.156 mrg 2582: * not follow links.
2583: */
2584: int
1.335 dsl 2585: sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval)
1.156 mrg 2586: {
1.335 dsl 2587: /* {
1.156 mrg 2588: syscallarg(const char *) path;
2589: syscallarg(u_long) flags;
1.335 dsl 2590: } */
1.163 enami 2591: struct vnode *vp;
1.156 mrg 2592: int error;
2593: struct nameidata nd;
2594:
1.331 pooka 2595: NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE,
1.334 pooka 2596: SCARG(uap, path));
1.156 mrg 2597: if ((error = namei(&nd)) != 0)
2598: return (error);
2599: vp = nd.ni_vp;
1.234 christos 2600: error = change_flags(vp, SCARG(uap, flags), l);
1.163 enami 2601: vput(vp);
2602: return (error);
2603: }
2604:
2605: /*
2606: * Common routine to change flags of a file.
2607: */
2608: int
1.234 christos 2609: change_flags(struct vnode *vp, u_long flags, struct lwp *l)
1.163 enami 2610: {
2611: struct vattr vattr;
2612: int error;
2613:
1.156 mrg 2614: vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1.163 enami 2615: /*
2616: * Non-superusers cannot change the flags on devices, even if they
2617: * own them.
2618: */
1.294 elad 2619: if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) {
1.332 pooka 2620: if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0)
1.156 mrg 2621: goto out;
2622: if (vattr.va_type == VCHR || vattr.va_type == VBLK) {
2623: error = EINVAL;
2624: goto out;
2625: }
2626: }
2627: VATTR_NULL(&vattr);
1.163 enami 2628: vattr.va_flags = flags;
1.332 pooka 2629: error = VOP_SETATTR(vp, &vattr, l->l_cred);
1.156 mrg 2630: out:
1.31 cgd 2631: return (error);
2632: }
2633:
2634: /*
1.98 enami 2635: * Change mode of a file given path name; this version follows links.
1.31 cgd 2636: */
2637: /* ARGSUSED */
1.63 christos 2638: int
1.335 dsl 2639: sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval)
1.56 thorpej 2640: {
1.335 dsl 2641: /* {
1.74 cgd 2642: syscallarg(const char *) path;
1.35 cgd 2643: syscallarg(int) mode;
1.335 dsl 2644: } */
1.31 cgd 2645: int error;
2646: struct nameidata nd;
2647:
1.331 pooka 2648: NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
1.334 pooka 2649: SCARG(uap, path));
1.63 christos 2650: if ((error = namei(&nd)) != 0)
1.31 cgd 2651: return (error);
1.97 enami 2652:
1.234 christos 2653: error = change_mode(nd.ni_vp, SCARG(uap, mode), l);
1.97 enami 2654:
2655: vrele(nd.ni_vp);
1.31 cgd 2656: return (error);
2657: }
2658:
2659: /*
2660: * Change mode of a file given a file descriptor.
2661: */
2662: /* ARGSUSED */
1.63 christos 2663: int
1.335 dsl 2664: sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval)
1.56 thorpej 2665: {
1.335 dsl 2666: /* {
1.35 cgd 2667: syscallarg(int) fd;
2668: syscallarg(int) mode;
1.335 dsl 2669: } */
1.346 ad 2670: file_t *fp;
1.31 cgd 2671: int error;
2672:
1.346 ad 2673: /* fd_getvnode() will use the descriptor for us */
2674: if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
1.31 cgd 2675: return (error);
1.346 ad 2676: error = change_mode(fp->f_data, SCARG(uap, mode), l);
2677: fd_putfile(SCARG(uap, fd));
1.135 thorpej 2678: return (error);
1.97 enami 2679: }
2680:
2681: /*
1.98 enami 2682: * Change mode of a file given path name; this version does not follow links.
2683: */
2684: /* ARGSUSED */
2685: int
1.335 dsl 2686: sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval)
1.98 enami 2687: {
1.335 dsl 2688: /* {
1.98 enami 2689: syscallarg(const char *) path;
2690: syscallarg(int) mode;
1.335 dsl 2691: } */
1.98 enami 2692: int error;
2693: struct nameidata nd;
2694:
1.331 pooka 2695: NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE,
1.334 pooka 2696: SCARG(uap, path));
1.98 enami 2697: if ((error = namei(&nd)) != 0)
2698: return (error);
2699:
1.234 christos 2700: error = change_mode(nd.ni_vp, SCARG(uap, mode), l);
1.98 enami 2701:
2702: vrele(nd.ni_vp);
2703: return (error);
2704: }
2705:
2706: /*
1.97 enami 2707: * Common routine to set mode given a vnode.
2708: */
2709: static int
1.234 christos 2710: change_mode(struct vnode *vp, int mode, struct lwp *l)
1.97 enami 2711: {
2712: struct vattr vattr;
2713: int error;
2714:
1.113 fvdl 2715: vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2716: VATTR_NULL(&vattr);
2717: vattr.va_mode = mode & ALLPERMS;
1.332 pooka 2718: error = VOP_SETATTR(vp, &vattr, l->l_cred);
1.113 fvdl 2719: VOP_UNLOCK(vp, 0);
1.31 cgd 2720: return (error);
2721: }
2722:
2723: /*
1.98 enami 2724: * Set ownership given a path name; this version follows links.
1.31 cgd 2725: */
2726: /* ARGSUSED */
1.63 christos 2727: int
1.335 dsl 2728: sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval)
1.56 thorpej 2729: {
1.335 dsl 2730: /* {
1.74 cgd 2731: syscallarg(const char *) path;
2732: syscallarg(uid_t) uid;
2733: syscallarg(gid_t) gid;
1.335 dsl 2734: } */
1.31 cgd 2735: int error;
2736: struct nameidata nd;
2737:
1.331 pooka 2738: NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
1.334 pooka 2739: SCARG(uap, path));
1.63 christos 2740: if ((error = namei(&nd)) != 0)
1.31 cgd 2741: return (error);
1.86 kleink 2742:
1.234 christos 2743: error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0);
1.112 kleink 2744:
2745: vrele(nd.ni_vp);
2746: return (error);
2747: }
2748:
2749: /*
2750: * Set ownership given a path name; this version follows links.
2751: * Provides POSIX semantics.
2752: */
2753: /* ARGSUSED */
2754: int
1.335 dsl 2755: sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval)
1.112 kleink 2756: {
1.335 dsl 2757: /* {
1.112 kleink 2758: syscallarg(const char *) path;
2759: syscallarg(uid_t) uid;
2760: syscallarg(gid_t) gid;
1.335 dsl 2761: } */
1.112 kleink 2762: int error;
2763: struct nameidata nd;
2764:
1.331 pooka 2765: NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
1.334 pooka 2766: SCARG(uap, path));
1.112 kleink 2767: if ((error = namei(&nd)) != 0)
2768: return (error);
2769:
1.234 christos 2770: error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1);
1.86 kleink 2771:
2772: vrele(nd.ni_vp);
1.31 cgd 2773: return (error);
2774: }
2775:
2776: /*
2777: * Set ownership given a file descriptor.
2778: */
2779: /* ARGSUSED */
1.63 christos 2780: int
1.335 dsl 2781: sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval)
1.56 thorpej 2782: {
1.335 dsl 2783: /* {
1.35 cgd 2784: syscallarg(int) fd;
1.74 cgd 2785: syscallarg(uid_t) uid;
2786: syscallarg(gid_t) gid;
1.335 dsl 2787: } */
1.71 mycroft 2788: int error;
1.346 ad 2789: file_t *fp;
1.31 cgd 2790:
1.346 ad 2791: /* fd_getvnode() will use the descriptor for us */
2792: if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
1.31 cgd 2793: return (error);
1.346 ad 2794: error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid),
2795: l, 0);
2796: fd_putfile(SCARG(uap, fd));
1.135 thorpej 2797: return (error);
1.112 kleink 2798: }
2799:
2800: /*
2801: * Set ownership given a file descriptor, providing POSIX/XPG semantics.
2802: */
2803: /* ARGSUSED */
2804: int
1.335 dsl 2805: sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval)
1.112 kleink 2806: {
1.335 dsl 2807: /* {
1.112 kleink 2808: syscallarg(int) fd;
2809: syscallarg(uid_t) uid;
2810: syscallarg(gid_t) gid;
1.335 dsl 2811: } */
1.112 kleink 2812: int error;
1.346 ad 2813: file_t *fp;
1.112 kleink 2814:
1.346 ad 2815: /* fd_getvnode() will use the descriptor for us */
2816: if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
1.112 kleink 2817: return (error);
1.346 ad 2818: error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid),
2819: l, 1);
2820: fd_putfile(SCARG(uap, fd));
1.135 thorpej 2821: return (error);
1.86 kleink 2822: }
2823:
2824: /*
1.98 enami 2825: * Set ownership given a path name; this version does not follow links.
2826: */
2827: /* ARGSUSED */
2828: int
1.335 dsl 2829: sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval)
1.98 enami 2830: {
1.335 dsl 2831: /* {
1.98 enami 2832: syscallarg(const char *) path;
2833: syscallarg(uid_t) uid;
2834: syscallarg(gid_t) gid;
1.335 dsl 2835: } */
1.98 enami 2836: int error;
2837: struct nameidata nd;
2838:
1.331 pooka 2839: NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE,
1.334 pooka 2840: SCARG(uap, path));
1.98 enami 2841: if ((error = namei(&nd)) != 0)
2842: return (error);
2843:
1.234 christos 2844: error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0);
1.112 kleink 2845:
2846: vrele(nd.ni_vp);
2847: return (error);
2848: }
2849:
2850: /*
2851: * Set ownership given a path name; this version does not follow links.
2852: * Provides POSIX/XPG semantics.
2853: */
2854: /* ARGSUSED */
2855: int
1.335 dsl 2856: sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval)
1.112 kleink 2857: {
1.335 dsl 2858: /* {
1.112 kleink 2859: syscallarg(const char *) path;
2860: syscallarg(uid_t) uid;
2861: syscallarg(gid_t) gid;
1.335 dsl 2862: } */
1.112 kleink 2863: int error;
2864: struct nameidata nd;
2865:
1.331 pooka 2866: NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE,
1.334 pooka 2867: SCARG(uap, path));
1.112 kleink 2868: if ((error = namei(&nd)) != 0)
2869: return (error);
2870:
1.234 christos 2871: error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1);
1.98 enami 2872:
2873: vrele(nd.ni_vp);
2874: return (error);
2875: }
2876:
2877: /*
1.205 junyoung 2878: * Common routine to set ownership given a vnode.
1.86 kleink 2879: */
2880: static int
1.234 christos 2881: change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l,
1.221 thorpej 2882: int posix_semantics)
1.86 kleink 2883: {
2884: struct vattr vattr;
1.112 kleink 2885: mode_t newmode;
1.86 kleink 2886: int error;
2887:
1.113 fvdl 2888: vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1.332 pooka 2889: if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0)
1.86 kleink 2890: goto out;
2891:
1.175 thorpej 2892: #define CHANGED(x) ((int)(x) != -1)
1.112 kleink 2893: newmode = vattr.va_mode;
2894: if (posix_semantics) {
2895: /*
1.114 kleink 2896: * POSIX/XPG semantics: if the caller is not the super-user,
2897: * clear set-user-id and set-group-id bits. Both POSIX and
2898: * the XPG consider the behaviour for calls by the super-user
2899: * implementation-defined; we leave the set-user-id and set-
2900: * group-id settings intact in that case.
1.112 kleink 2901: */
1.257 ad 2902: if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER,
1.242 elad 2903: NULL) != 0)
1.112 kleink 2904: newmode &= ~(S_ISUID | S_ISGID);
2905: } else {
2906: /*
2907: * NetBSD semantics: when changing owner and/or group,
2908: * clear the respective bit(s).
2909: */
2910: if (CHANGED(uid))
2911: newmode &= ~S_ISUID;
2912: if (CHANGED(gid))
2913: newmode &= ~S_ISGID;
2914: }
2915: /* Update va_mode iff altered. */
2916: if (vattr.va_mode == newmode)
2917: newmode = VNOVAL;
1.205 junyoung 2918:
1.86 kleink 2919: VATTR_NULL(&vattr);
1.175 thorpej 2920: vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL;
2921: vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL;
1.86 kleink 2922: vattr.va_mode = newmode;
1.332 pooka 2923: error = VOP_SETATTR(vp, &vattr, l->l_cred);
1.112 kleink 2924: #undef CHANGED
1.205 junyoung 2925:
1.86 kleink 2926: out:
1.113 fvdl 2927: VOP_UNLOCK(vp, 0);
1.31 cgd 2928: return (error);
2929: }
2930:
2931: /*
1.98 enami 2932: * Set the access and modification times given a path name; this
2933: * version follows links.
1.31 cgd 2934: */
2935: /* ARGSUSED */
1.63 christos 2936: int
1.383 christos 2937: sys___utimes50(struct lwp *l, const struct sys___utimes50_args *uap,
2938: register_t *retval)
1.56 thorpej 2939: {
1.335 dsl 2940: /* {
1.74 cgd 2941: syscallarg(const char *) path;
2942: syscallarg(const struct timeval *) tptr;
1.335 dsl 2943: } */
1.31 cgd 2944:
1.312 dsl 2945: return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW,
1.346 ad 2946: SCARG(uap, tptr), UIO_USERSPACE);
1.71 mycroft 2947: }
2948:
2949: /*
2950: * Set the access and modification times given a file descriptor.
2951: */
2952: /* ARGSUSED */
2953: int
1.383 christos 2954: sys___futimes50(struct lwp *l, const struct sys___futimes50_args *uap,
2955: register_t *retval)
1.71 mycroft 2956: {
1.335 dsl 2957: /* {
1.71 mycroft 2958: syscallarg(int) fd;
1.74 cgd 2959: syscallarg(const struct timeval *) tptr;
1.335 dsl 2960: } */
1.71 mycroft 2961: int error;
1.346 ad 2962: file_t *fp;
1.71 mycroft 2963:
1.346 ad 2964: /* fd_getvnode() will use the descriptor for us */
2965: if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
1.96 enami 2966: return (error);
1.346 ad 2967: error = do_sys_utimes(l, fp->f_data, NULL, 0, SCARG(uap, tptr),
2968: UIO_USERSPACE);
2969: fd_putfile(SCARG(uap, fd));
1.135 thorpej 2970: return (error);
1.98 enami 2971: }
2972:
2973: /*
2974: * Set the access and modification times given a path name; this
2975: * version does not follow links.
2976: */
2977: int
1.383 christos 2978: sys___lutimes50(struct lwp *l, const struct sys___lutimes50_args *uap,
2979: register_t *retval)
1.98 enami 2980: {
1.335 dsl 2981: /* {
1.98 enami 2982: syscallarg(const char *) path;
2983: syscallarg(const struct timeval *) tptr;
1.335 dsl 2984: } */
1.98 enami 2985:
1.312 dsl 2986: return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW,
1.346 ad 2987: SCARG(uap, tptr), UIO_USERSPACE);
1.97 enami 2988: }
2989:
2990: /*
2991: * Common routine to set access and modification times given a vnode.
2992: */
1.312 dsl 2993: int
2994: do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag,
2995: const struct timeval *tptr, enum uio_seg seg)
1.97 enami 2996: {
2997: struct vattr vattr;
1.312 dsl 2998: struct nameidata nd;
1.97 enami 2999: int error;
1.367 christos 3000: bool vanull, setbirthtime;
3001: struct timespec ts[2];
1.97 enami 3002:
3003: if (tptr == NULL) {
1.367 christos 3004: vanull = true;
3005: nanotime(&ts[0]);
3006: ts[1] = ts[0];
1.71 mycroft 3007: } else {
1.233 yamt 3008: struct timeval tv[2];
3009:
1.367 christos 3010: vanull = false;
1.312 dsl 3011: if (seg != UIO_SYSSPACE) {
1.383 christos 3012: error = copyin(tptr, tv, sizeof (tv));
1.312 dsl 3013: if (error != 0)
3014: return error;
3015: tptr = tv;
3016: }
1.367 christos 3017: TIMEVAL_TO_TIMESPEC(&tptr[0], &ts[0]);
3018: TIMEVAL_TO_TIMESPEC(&tptr[1], &ts[1]);
1.71 mycroft 3019: }
1.312 dsl 3020:
3021: if (vp == NULL) {
1.334 pooka 3022: NDINIT(&nd, LOOKUP, flag | TRYEMULROOT, UIO_USERSPACE, path);
1.312 dsl 3023: if ((error = namei(&nd)) != 0)
1.367 christos 3024: return error;
1.312 dsl 3025: vp = nd.ni_vp;
3026: } else
3027: nd.ni_vp = NULL;
3028:
1.113 fvdl 3029: vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1.367 christos 3030: setbirthtime = (VOP_GETATTR(vp, &vattr, l->l_cred) == 0 &&
3031: timespeccmp(&ts[1], &vattr.va_birthtime, <));
3032: VATTR_NULL(&vattr);
1.368 christos 3033: vattr.va_atime = ts[0];
3034: vattr.va_mtime = ts[1];
1.367 christos 3035: if (setbirthtime)
3036: vattr.va_birthtime = ts[1];
3037: if (vanull)
3038: vattr.va_flags |= VA_UTIMES_NULL;
1.332 pooka 3039: error = VOP_SETATTR(vp, &vattr, l->l_cred);
1.113 fvdl 3040: VOP_UNLOCK(vp, 0);
1.312 dsl 3041:
3042: if (nd.ni_vp != NULL)
3043: vrele(nd.ni_vp);
3044:
1.367 christos 3045: return error;
1.31 cgd 3046: }
3047:
3048: /*
3049: * Truncate a file given its path name.
3050: */
3051: /* ARGSUSED */
1.63 christos 3052: int
1.335 dsl 3053: sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval)
1.56 thorpej 3054: {
1.335 dsl 3055: /* {
1.74 cgd 3056: syscallarg(const char *) path;
1.35 cgd 3057: syscallarg(int) pad;
3058: syscallarg(off_t) length;
1.335 dsl 3059: } */
1.155 augustss 3060: struct vnode *vp;
1.31 cgd 3061: struct vattr vattr;
3062: int error;
3063: struct nameidata nd;
3064:
1.331 pooka 3065: NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
1.334 pooka 3066: SCARG(uap, path));
1.63 christos 3067: if ((error = namei(&nd)) != 0)
1.31 cgd 3068: return (error);
3069: vp = nd.ni_vp;
1.113 fvdl 3070: vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1.31 cgd 3071: if (vp->v_type == VDIR)
3072: error = EISDIR;
3073: else if ((error = vn_writechk(vp)) == 0 &&
1.332 pooka 3074: (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) {
1.31 cgd 3075: VATTR_NULL(&vattr);
1.35 cgd 3076: vattr.va_size = SCARG(uap, length);
1.332 pooka 3077: error = VOP_SETATTR(vp, &vattr, l->l_cred);
1.31 cgd 3078: }
3079: vput(vp);
3080: return (error);
3081: }
3082:
3083: /*
3084: * Truncate a file given a file descriptor.
3085: */
3086: /* ARGSUSED */
1.63 christos 3087: int
1.335 dsl 3088: sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval)
1.56 thorpej 3089: {
1.335 dsl 3090: /* {
1.35 cgd 3091: syscallarg(int) fd;
3092: syscallarg(int) pad;
3093: syscallarg(off_t) length;
1.335 dsl 3094: } */
1.31 cgd 3095: struct vattr vattr;
3096: struct vnode *vp;
1.346 ad 3097: file_t *fp;
1.31 cgd 3098: int error;
3099:
1.346 ad 3100: /* fd_getvnode() will use the descriptor for us */
3101: if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
1.31 cgd 3102: return (error);
1.135 thorpej 3103: if ((fp->f_flag & FWRITE) == 0) {
3104: error = EINVAL;
3105: goto out;
3106: }
1.346 ad 3107: vp = fp->f_data;
1.113 fvdl 3108: vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1.31 cgd 3109: if (vp->v_type == VDIR)
3110: error = EISDIR;
3111: else if ((error = vn_writechk(vp)) == 0) {
3112: VATTR_NULL(&vattr);
1.35 cgd 3113: vattr.va_size = SCARG(uap, length);
1.332 pooka 3114: error = VOP_SETATTR(vp, &vattr, fp->f_cred);
1.31 cgd 3115: }
1.113 fvdl 3116: VOP_UNLOCK(vp, 0);
1.135 thorpej 3117: out:
1.346 ad 3118: fd_putfile(SCARG(uap, fd));
1.31 cgd 3119: return (error);
3120: }
3121:
3122: /*
3123: * Sync an open file.
3124: */
3125: /* ARGSUSED */
1.63 christos 3126: int
1.335 dsl 3127: sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval)
1.56 thorpej 3128: {
1.335 dsl 3129: /* {
1.35 cgd 3130: syscallarg(int) fd;
1.335 dsl 3131: } */
1.155 augustss 3132: struct vnode *vp;
1.346 ad 3133: file_t *fp;
1.31 cgd 3134: int error;
3135:
1.346 ad 3136: /* fd_getvnode() will use the descriptor for us */
3137: if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
1.31 cgd 3138: return (error);
1.346 ad 3139: vp = fp->f_data;
1.113 fvdl 3140: vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1.332 pooka 3141: error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0);
1.113 fvdl 3142: VOP_UNLOCK(vp, 0);
1.346 ad 3143: fd_putfile(SCARG(uap, fd));
1.201 thorpej 3144: return (error);
3145: }
3146:
3147: /*
3148: * Sync a range of file data. API modeled after that found in AIX.
3149: *
3150: * FDATASYNC indicates that we need only save enough metadata to be able
3151: * to re-read the written data. Note we duplicate AIX's requirement that
3152: * the file be open for writing.
3153: */
3154: /* ARGSUSED */
3155: int
1.335 dsl 3156: sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval)
1.201 thorpej 3157: {
1.335 dsl 3158: /* {
1.201 thorpej 3159: syscallarg(int) fd;
3160: syscallarg(int) flags;
3161: syscallarg(off_t) start;
1.225 cube 3162: syscallarg(off_t) length;
1.335 dsl 3163: } */
1.201 thorpej 3164: struct vnode *vp;
1.346 ad 3165: file_t *fp;
1.201 thorpej 3166: int flags, nflags;
3167: off_t s, e, len;
3168: int error;
3169:
1.346 ad 3170: /* fd_getvnode() will use the descriptor for us */
3171: if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
1.201 thorpej 3172: return (error);
3173:
3174: if ((fp->f_flag & FWRITE) == 0) {
1.293 wrstuden 3175: error = EBADF;
3176: goto out;
1.201 thorpej 3177: }
3178:
3179: flags = SCARG(uap, flags);
3180: if (((flags & (FDATASYNC | FFILESYNC)) == 0) ||
3181: ((~flags & (FDATASYNC | FFILESYNC)) == 0)) {
1.293 wrstuden 3182: error = EINVAL;
3183: goto out;
1.201 thorpej 3184: }
3185: /* Now set up the flags for value(s) to pass to VOP_FSYNC() */
3186: if (flags & FDATASYNC)
3187: nflags = FSYNC_DATAONLY | FSYNC_WAIT;
3188: else
3189: nflags = FSYNC_WAIT;
1.216 wrstuden 3190: if (flags & FDISKSYNC)
3191: nflags |= FSYNC_CACHE;
1.201 thorpej 3192:
3193: len = SCARG(uap, length);
3194: /* If length == 0, we do the whole file, and s = l = 0 will do that */
3195: if (len) {
3196: s = SCARG(uap, start);
3197: e = s + len;
3198: if (e < s) {
1.293 wrstuden 3199: error = EINVAL;
3200: goto out;
1.201 thorpej 3201: }
3202: } else {
3203: e = 0;
3204: s = 0;
3205: }
3206:
1.346 ad 3207: vp = fp->f_data;
1.201 thorpej 3208: vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1.332 pooka 3209: error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e);
1.201 thorpej 3210: VOP_UNLOCK(vp, 0);
1.293 wrstuden 3211: out:
1.346 ad 3212: fd_putfile(SCARG(uap, fd));
1.31 cgd 3213: return (error);
3214: }
3215:
3216: /*
1.117 kleink 3217: * Sync the data of an open file.
3218: */
3219: /* ARGSUSED */
3220: int
1.335 dsl 3221: sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval)
1.117 kleink 3222: {
1.335 dsl 3223: /* {
1.117 kleink 3224: syscallarg(int) fd;
1.335 dsl 3225: } */
1.117 kleink 3226: struct vnode *vp;
1.346 ad 3227: file_t *fp;
1.117 kleink 3228: int error;
3229:
1.346 ad 3230: /* fd_getvnode() will use the descriptor for us */
3231: if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
1.117 kleink 3232: return (error);
1.199 kleink 3233: if ((fp->f_flag & FWRITE) == 0) {
1.346 ad 3234: fd_putfile(SCARG(uap, fd));
1.199 kleink 3235: return (EBADF);
3236: }
1.346 ad 3237: vp = fp->f_data;
1.117 kleink 3238: vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1.332 pooka 3239: error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0);
1.117 kleink 3240: VOP_UNLOCK(vp, 0);
1.346 ad 3241: fd_putfile(SCARG(uap, fd));
1.117 kleink 3242: return (error);
3243: }
3244:
3245: /*
1.90 kleink 3246: * Rename files, (standard) BSD semantics frontend.
1.31 cgd 3247: */
3248: /* ARGSUSED */
1.63 christos 3249: int
1.335 dsl 3250: sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval)
1.56 thorpej 3251: {
1.335 dsl 3252: /* {
1.74 cgd 3253: syscallarg(const char *) from;
3254: syscallarg(const char *) to;
1.335 dsl 3255: } */
1.90 kleink 3256:
1.336 ad 3257: return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 0));
1.90 kleink 3258: }
3259:
3260: /*
3261: * Rename files, POSIX semantics frontend.
3262: */
3263: /* ARGSUSED */
3264: int
1.335 dsl 3265: sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval)
1.90 kleink 3266: {
1.335 dsl 3267: /* {
1.90 kleink 3268: syscallarg(const char *) from;
3269: syscallarg(const char *) to;
1.335 dsl 3270: } */
1.90 kleink 3271:
1.336 ad 3272: return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 1));
1.90 kleink 3273: }
3274:
3275: /*
3276: * Rename files. Source and destination must either both be directories,
3277: * or both not be directories. If target is a directory, it must be empty.
3278: * If `from' and `to' refer to the same object, the value of the `retain'
3279: * argument is used to determine whether `from' will be
3280: *
3281: * (retain == 0) deleted unless `from' and `to' refer to the same
3282: * object in the file system's name space (BSD).
3283: * (retain == 1) always retained (POSIX).
3284: */
1.336 ad 3285: int
3286: do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain)
1.90 kleink 3287: {
1.155 augustss 3288: struct vnode *tvp, *fvp, *tdvp;
1.31 cgd 3289: struct nameidata fromnd, tond;
1.344 dholland 3290: struct mount *fs;
1.336 ad 3291: struct lwp *l = curlwp;
1.234 christos 3292: struct proc *p;
1.344 dholland 3293: uint32_t saveflag;
1.31 cgd 3294: int error;
3295:
1.331 pooka 3296: NDINIT(&fromnd, DELETE, LOCKPARENT | SAVESTART | TRYEMULROOT,
1.336 ad 3297: seg, from);
1.63 christos 3298: if ((error = namei(&fromnd)) != 0)
1.31 cgd 3299: return (error);
1.291 pooka 3300: if (fromnd.ni_dvp != fromnd.ni_vp)
3301: VOP_UNLOCK(fromnd.ni_dvp, 0);
1.31 cgd 3302: fvp = fromnd.ni_vp;
1.344 dholland 3303:
3304: fs = fvp->v_mount;
3305: error = VFS_RENAMELOCK_ENTER(fs);
3306: if (error) {
3307: VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3308: vrele(fromnd.ni_dvp);
3309: vrele(fvp);
3310: goto out1;
3311: }
3312:
3313: /*
3314: * close, partially, yet another race - ideally we should only
3315: * go as far as getting fromnd.ni_dvp before getting the per-fs
3316: * lock, and then continue to get fromnd.ni_vp, but we can't do
3317: * that with namei as it stands.
3318: *
3319: * This still won't prevent rmdir from nuking fromnd.ni_vp
3320: * under us. The real fix is to get the locks in the right
3321: * order and do the lookups in the right places, but that's a
3322: * major rototill.
3323: *
3324: * Preserve the SAVESTART in cn_flags, because who knows what
3325: * might happen if we don't.
3326: *
3327: * Note: this logic (as well as this whole function) is cloned
3328: * in nfs_serv.c. Proceed accordingly.
3329: */
3330: vrele(fvp);
1.348 dholland 3331: if ((fromnd.ni_cnd.cn_namelen == 1 &&
3332: fromnd.ni_cnd.cn_nameptr[0] == '.') ||
3333: (fromnd.ni_cnd.cn_namelen == 2 &&
3334: fromnd.ni_cnd.cn_nameptr[0] == '.' &&
3335: fromnd.ni_cnd.cn_nameptr[1] == '.')) {
3336: error = EINVAL;
3337: VFS_RENAMELOCK_EXIT(fs);
3338: VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3339: vrele(fromnd.ni_dvp);
3340: goto out1;
3341: }
1.344 dholland 3342: saveflag = fromnd.ni_cnd.cn_flags & SAVESTART;
3343: fromnd.ni_cnd.cn_flags &= ~SAVESTART;
3344: vn_lock(fromnd.ni_dvp, LK_EXCLUSIVE | LK_RETRY);
3345: error = relookup(fromnd.ni_dvp, &fromnd.ni_vp, &fromnd.ni_cnd);
3346: fromnd.ni_cnd.cn_flags |= saveflag;
3347: if (error) {
3348: VOP_UNLOCK(fromnd.ni_dvp, 0);
3349: VFS_RENAMELOCK_EXIT(fs);
3350: VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3351: vrele(fromnd.ni_dvp);
3352: goto out1;
3353: }
3354: VOP_UNLOCK(fromnd.ni_vp, 0);
3355: if (fromnd.ni_dvp != fromnd.ni_vp)
3356: VOP_UNLOCK(fromnd.ni_dvp, 0);
3357: fvp = fromnd.ni_vp;
3358:
1.331 pooka 3359: NDINIT(&tond, RENAME,
3360: LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | TRYEMULROOT
3361: | (fvp->v_type == VDIR ? CREATEDIR : 0),
1.336 ad 3362: seg, to);
1.63 christos 3363: if ((error = namei(&tond)) != 0) {
1.344 dholland 3364: VFS_RENAMELOCK_EXIT(fs);
1.31 cgd 3365: VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3366: vrele(fromnd.ni_dvp);
3367: vrele(fvp);
3368: goto out1;
3369: }
3370: tdvp = tond.ni_dvp;
3371: tvp = tond.ni_vp;
1.90 kleink 3372:
1.31 cgd 3373: if (tvp != NULL) {
3374: if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3375: error = ENOTDIR;
3376: goto out;
3377: } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3378: error = EISDIR;
3379: goto out;
3380: }
3381: }
1.90 kleink 3382:
1.31 cgd 3383: if (fvp == tdvp)
3384: error = EINVAL;
1.90 kleink 3385:
1.82 kleink 3386: /*
1.90 kleink 3387: * Source and destination refer to the same object.
1.82 kleink 3388: */
1.90 kleink 3389: if (fvp == tvp) {
3390: if (retain)
3391: error = -1;
3392: else if (fromnd.ni_dvp == tdvp &&
3393: fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen &&
1.123 perry 3394: !memcmp(fromnd.ni_cnd.cn_nameptr,
1.90 kleink 3395: tond.ni_cnd.cn_nameptr,
3396: fromnd.ni_cnd.cn_namelen))
1.82 kleink 3397: error = -1;
1.90 kleink 3398: }
3399:
1.256 elad 3400: #if NVERIEXEC > 0
1.282 elad 3401: if (!error) {
1.313 elad 3402: char *f1, *f2;
1.384 yamt 3403: size_t f1_len;
3404: size_t f2_len;
1.313 elad 3405:
1.384 yamt 3406: f1_len = fromnd.ni_cnd.cn_namelen + 1;
3407: f1 = kmem_alloc(f1_len, KM_SLEEP);
3408: strlcpy(f1, fromnd.ni_cnd.cn_nameptr, f1_len);
3409:
3410: f2_len = tond.ni_cnd.cn_namelen + 1;
3411: f2 = kmem_alloc(f2_len, KM_SLEEP);
3412: strlcpy(f2, tond.ni_cnd.cn_nameptr, f2_len);
1.282 elad 3413:
1.315 christos 3414: error = veriexec_renamechk(l, fvp, f1, tvp, f2);
1.282 elad 3415:
1.384 yamt 3416: kmem_free(f1, f1_len);
3417: kmem_free(f2, f2_len);
1.282 elad 3418: }
1.256 elad 3419: #endif /* NVERIEXEC > 0 */
1.229 elad 3420:
1.31 cgd 3421: out:
1.234 christos 3422: p = l->l_proc;
1.31 cgd 3423: if (!error) {
3424: error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3425: tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
1.344 dholland 3426: VFS_RENAMELOCK_EXIT(fs);
1.31 cgd 3427: } else {
3428: VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
3429: if (tdvp == tvp)
3430: vrele(tdvp);
3431: else
3432: vput(tdvp);
3433: if (tvp)
3434: vput(tvp);
1.344 dholland 3435: VFS_RENAMELOCK_EXIT(fs);
1.31 cgd 3436: VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3437: vrele(fromnd.ni_dvp);
3438: vrele(fvp);
3439: }
3440: vrele(tond.ni_startdir);
1.161 thorpej 3441: PNBUF_PUT(tond.ni_cnd.cn_pnbuf);
1.31 cgd 3442: out1:
3443: if (fromnd.ni_startdir)
3444: vrele(fromnd.ni_startdir);
1.161 thorpej 3445: PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf);
1.80 kleink 3446: return (error == -1 ? 0 : error);
1.31 cgd 3447: }
3448:
3449: /*
3450: * Make a directory file.
3451: */
3452: /* ARGSUSED */
1.63 christos 3453: int
1.335 dsl 3454: sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval)
1.56 thorpej 3455: {
1.335 dsl 3456: /* {
1.74 cgd 3457: syscallarg(const char *) path;
1.35 cgd 3458: syscallarg(int) mode;
1.335 dsl 3459: } */
1.179 thorpej 3460: struct proc *p = l->l_proc;
1.155 augustss 3461: struct vnode *vp;
1.31 cgd 3462: struct vattr vattr;
3463: int error;
3464: struct nameidata nd;
3465:
1.310 dsl 3466: NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, UIO_USERSPACE,
1.334 pooka 3467: SCARG(uap, path));
1.63 christos 3468: if ((error = namei(&nd)) != 0)
1.31 cgd 3469: return (error);
3470: vp = nd.ni_vp;
3471: if (vp != NULL) {
3472: VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3473: if (nd.ni_dvp == vp)
3474: vrele(nd.ni_dvp);
3475: else
3476: vput(nd.ni_dvp);
3477: vrele(vp);
3478: return (EEXIST);
3479: }
3480: VATTR_NULL(&vattr);
3481: vattr.va_type = VDIR;
1.328 ad 3482: /* We will read cwdi->cwdi_cmask unlocked. */
1.134 thorpej 3483: vattr.va_mode =
3484: (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask;
1.31 cgd 3485: error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3486: if (!error)
3487: vput(nd.ni_vp);
3488: return (error);
3489: }
3490:
3491: /*
3492: * Remove a directory file.
3493: */
3494: /* ARGSUSED */
1.63 christos 3495: int
1.335 dsl 3496: sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval)
1.56 thorpej 3497: {
1.335 dsl 3498: /* {
1.74 cgd 3499: syscallarg(const char *) path;
1.335 dsl 3500: } */
1.155 augustss 3501: struct vnode *vp;
1.31 cgd 3502: int error;
3503: struct nameidata nd;
3504:
1.310 dsl 3505: NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
1.334 pooka 3506: SCARG(uap, path));
1.63 christos 3507: if ((error = namei(&nd)) != 0)
1.31 cgd 3508: return (error);
3509: vp = nd.ni_vp;
3510: if (vp->v_type != VDIR) {
3511: error = ENOTDIR;
3512: goto out;
3513: }
3514: /*
3515: * No rmdir "." please.
3516: */
3517: if (nd.ni_dvp == vp) {
3518: error = EINVAL;
3519: goto out;
3520: }
3521: /*
3522: * The root of a mounted filesystem cannot be deleted.
3523: */
1.350 joerg 3524: if ((vp->v_vflag & VV_ROOT) != 0 || vp->v_mountedhere != NULL) {
1.31 cgd 3525: error = EBUSY;
1.197 hannken 3526: goto out;
3527: }
3528: error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3529: return (error);
3530:
3531: out:
3532: VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3533: if (nd.ni_dvp == vp)
3534: vrele(nd.ni_dvp);
3535: else
3536: vput(nd.ni_dvp);
3537: vput(vp);
1.31 cgd 3538: return (error);
3539: }
3540:
3541: /*
3542: * Read a block of directory entries in a file system independent format.
3543: */
1.63 christos 3544: int
1.335 dsl 3545: sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval)
1.56 thorpej 3546: {
1.335 dsl 3547: /* {
1.35 cgd 3548: syscallarg(int) fd;
3549: syscallarg(char *) buf;
1.101 fvdl 3550: syscallarg(size_t) count;
1.335 dsl 3551: } */
1.346 ad 3552: file_t *fp;
1.101 fvdl 3553: int error, done;
1.31 cgd 3554:
1.346 ad 3555: /* fd_getvnode() will use the descriptor for us */
3556: if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
1.31 cgd 3557: return (error);
1.135 thorpej 3558: if ((fp->f_flag & FREAD) == 0) {
3559: error = EBADF;
3560: goto out;
3561: }
1.101 fvdl 3562: error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE,
1.234 christos 3563: SCARG(uap, count), &done, l, 0, 0);
1.325 ad 3564: ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error);
1.101 fvdl 3565: *retval = done;
1.135 thorpej 3566: out:
1.346 ad 3567: fd_putfile(SCARG(uap, fd));
1.31 cgd 3568: return (error);
3569: }
3570:
3571: /*
3572: * Set the mode mask for creation of filesystem nodes.
3573: */
1.56 thorpej 3574: int
1.335 dsl 3575: sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval)
1.56 thorpej 3576: {
1.335 dsl 3577: /* {
1.103 mycroft 3578: syscallarg(mode_t) newmask;
1.335 dsl 3579: } */
1.179 thorpej 3580: struct proc *p = l->l_proc;
1.134 thorpej 3581: struct cwdinfo *cwdi;
1.31 cgd 3582:
1.328 ad 3583: /*
3584: * cwdi->cwdi_cmask will be read unlocked elsewhere. What's
3585: * important is that we serialize changes to the mask. The
3586: * rw_exit() will issue a write memory barrier on our behalf,
3587: * and force the changes out to other CPUs (as it must use an
3588: * atomic operation, draining the local CPU's store buffers).
3589: */
1.134 thorpej 3590: cwdi = p->p_cwdi;
1.328 ad 3591: rw_enter(&cwdi->cwdi_lock, RW_WRITER);
1.134 thorpej 3592: *retval = cwdi->cwdi_cmask;
3593: cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS;
1.328 ad 3594: rw_exit(&cwdi->cwdi_lock);
3595:
1.31 cgd 3596: return (0);
3597: }
3598:
1.340 elad 3599: int
3600: dorevoke(struct vnode *vp, kauth_cred_t cred)
3601: {
3602: struct vattr vattr;
3603: int error;
3604:
3605: if ((error = VOP_GETATTR(vp, &vattr, cred)) != 0)
1.342 ad 3606: return error;
1.385 enami 3607: if (kauth_cred_geteuid(cred) == vattr.va_uid ||
1.340 elad 3608: (error = kauth_authorize_generic(cred,
1.342 ad 3609: KAUTH_GENERIC_ISSUSER, NULL)) == 0)
1.340 elad 3610: VOP_REVOKE(vp, REVOKEALL);
3611: return (error);
3612: }
3613:
1.31 cgd 3614: /*
3615: * Void all references to file by ripping underlying filesystem
3616: * away from vnode.
3617: */
3618: /* ARGSUSED */
1.63 christos 3619: int
1.335 dsl 3620: sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval)
1.56 thorpej 3621: {
1.335 dsl 3622: /* {
1.74 cgd 3623: syscallarg(const char *) path;
1.335 dsl 3624: } */
1.155 augustss 3625: struct vnode *vp;
1.31 cgd 3626: int error;
3627: struct nameidata nd;
3628:
1.331 pooka 3629: NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
1.334 pooka 3630: SCARG(uap, path));
1.63 christos 3631: if ((error = namei(&nd)) != 0)
1.31 cgd 3632: return (error);
3633: vp = nd.ni_vp;
1.340 elad 3634: error = dorevoke(vp, l->l_cred);
1.31 cgd 3635: vrele(vp);
3636: return (error);
3637: }
CVSweb <webmaster@jp.NetBSD.org>
![[BACK]](/icons/back.gif){kind=icon}
Return to vfs_syscalls.c CVS log ![[TXT]](/icons/text.gif){kind=icon}
![[DIR]](/icons/dir.gif){kind=icon}
Up to [cvs.NetBSD.org] / src / sys / kern