Sneed-Group/gems-kernel
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
gems-kernel/source/THIRDPARTY/xnu/bsd/vfs/vfs_syscalls.c

14702 lines
359 KiB
C
Raw Normal View History

add darwin/xnu functionality
2024-06-03 11:29:39 -05:00
/*
* Copyright (c) 1995-2022 Apple Inc. All rights reserved.
*
* @APPLE_OSREFERENCE_LICENSE_HEADER_START@
*
* This file contains Original Code and/or Modifications of Original Code
* as defined in and that are subject to the Apple Public Source License
* Version 2.0 (the 'License'). You may not use this file except in
* compliance with the License. The rights granted to you under the License
* may not be used to create, or enable the creation or redistribution of,
* unlawful or unlicensed copies of an Apple operating system, or to
* circumvent, violate, or enable the circumvention or violation of, any
* terms of an Apple operating system software license agreement.
*
* Please obtain a copy of the License at
* http://www.opensource.apple.com/apsl/ and read it before using this file.
*
* The Original Code and all software distributed under the License are
* distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
* EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
* INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
* Please see the License for the specific language governing rights and
* limitations under the License.
*
* @APPLE_OSREFERENCE_LICENSE_HEADER_END@
*/
/*
* Copyright (c) 1989, 1993
* The Regents of the University of California. All rights reserved.
* (c) UNIX System Laboratories, Inc.
* All or some portions of this file are derived from material licensed
* to the University of California by American Telephone and Telegraph
* Co. or Unix System Laboratories, Inc. and are reproduced herein with
* the permission of UNIX System Laboratories, Inc.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by the University of
* California, Berkeley and its contributors.
* 4. Neither the name of the University nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
* @(#)vfs_syscalls.c 8.41 (Berkeley) 6/15/95
*/
/*
* NOTICE: This file was modified by SPARTA, Inc. in 2005 to introduce
* support for mandatory and extensible security protections. This notice
* is included in support of clause 2.2 (b) of the Apple Public License,
* Version 2.0.
*/
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/namei.h>
#include <sys/filedesc.h>
#include <sys/kernel.h>
#include <sys/file_internal.h>
#include <sys/stat.h>
#include <sys/vnode_internal.h>
#include <sys/mount_internal.h>
#include <sys/proc_internal.h>
#include <sys/kauth.h>
#include <sys/uio_internal.h>
#include <kern/kalloc.h>
#include <sys/mman.h>
#include <sys/dirent.h>
#include <sys/attr.h>
#include <sys/sysctl.h>
#include <sys/ubc.h>
#include <sys/quota.h>
#include <sys/kdebug.h>
#include <sys/fsevents.h>
#include <sys/imgsrc.h>
#include <sys/sysproto.h>
#include <sys/sysctl.h>
#include <sys/xattr.h>
#include <sys/fcntl.h>
#include <sys/stdio.h>
#include <sys/fsctl.h>
#include <sys/ubc_internal.h>
#include <sys/disk.h>
#include <sys/content_protection.h>
#include <sys/clonefile.h>
#include <sys/snapshot.h>
#include <sys/priv.h>
#include <sys/fsgetpath.h>
#include <machine/cons.h>
#include <machine/limits.h>
#include <miscfs/specfs/specdev.h>
#include <vfs/vfs_disk_conditioner.h>
#if CONFIG_EXCLAVES
#include <vfs/vfs_exclave_fs.h>
#endif
#include <security/audit/audit.h>
#include <bsm/audit_kevents.h>
#include <mach/mach_types.h>
#include <kern/kern_types.h>
#include <kern/kalloc.h>
#include <kern/task.h>
#include <vm/vm_pageout.h>
#include <vm/vm_protos.h>
#include <libkern/OSAtomic.h>
#include <os/atomic_private.h>
#include <pexpert/pexpert.h>
#include <IOKit/IOBSD.h>
// deps for MIG call
#include <kern/host.h>
#include <kern/ipc_misc.h>
#include <mach/host_priv.h>
#include <mach/vfs_nspace.h>
#include <os/log.h>
#include <nfs/nfs_conf.h>
#if ROUTEFS
#include <miscfs/routefs/routefs.h>
#endif /* ROUTEFS */
#if CONFIG_MACF
#include <security/mac.h>
#include <security/mac_framework.h>
#endif
#if CONFIG_FSE
#define GET_PATH(x) \
((x) = get_pathbuff())
#define RELEASE_PATH(x) \
release_pathbuff(x)
#else
#define GET_PATH(x) \
((x) = zalloc(ZV_NAMEI))
#define RELEASE_PATH(x) \
zfree(ZV_NAMEI, x)
#endif /* CONFIG_FSE */
#ifndef HFS_GET_BOOT_INFO
#define HFS_GET_BOOT_INFO (FCNTL_FS_SPECIFIC_BASE + 0x00004)
#endif
#ifndef HFS_SET_BOOT_INFO
#define HFS_SET_BOOT_INFO (FCNTL_FS_SPECIFIC_BASE + 0x00005)
#endif
#ifndef APFSIOC_REVERT_TO_SNAPSHOT
#define APFSIOC_REVERT_TO_SNAPSHOT _IOW('J', 1, u_int64_t)
#endif
extern void disk_conditioner_unmount(mount_t mp);
/* struct for checkdirs iteration */
struct cdirargs {
vnode_t olddp;
vnode_t newdp;
};
/* callback for checkdirs iteration */
static int checkdirs_callback(proc_t p, void * arg);
static int change_dir(struct nameidata *ndp, vfs_context_t ctx);
static int checkdirs(vnode_t olddp, vfs_context_t ctx);
void enablequotas(struct mount *mp, vfs_context_t ctx);
static int getfsstat_callback(mount_t mp, void * arg);
static int getutimes(user_addr_t usrtvp, struct timespec *tsp);
static int setutimes(vfs_context_t ctx, vnode_t vp, const struct timespec *ts, int nullflag);
static int sync_callback(mount_t, void *);
static int munge_statfs(struct mount *mp, struct vfsstatfs *sfsp,
user_addr_t bufp, int *sizep, boolean_t is_64_bit,
boolean_t partial_copy);
static int fsync_common(proc_t p, struct fsync_args *uap, int flags);
static int mount_common(const char *fstypename, vnode_t pvp, vnode_t vp,
struct componentname *cnp, user_addr_t fsmountargs,
int flags, uint32_t internal_flags, char *labelstr, vfs_context_t ctx);
void vfs_notify_mount(vnode_t pdvp);
int prepare_coveredvp(vnode_t vp, vfs_context_t ctx, struct componentname *cnp, const char *fsname, uint32_t internal_flags);
struct fd_vn_data * fg_vn_data_alloc(void);
/*
* Max retries for ENOENT returns from vn_authorize_{rmdir, unlink, rename}
* Concurrent lookups (or lookups by ids) on hard links can cause the
* vn_getpath (which does not re-enter the filesystem as vn_getpath_fsenter
* does) to return ENOENT as the path cannot be returned from the name cache
* alone. We have no option but to retry and hope to get one namei->reverse path
* generation done without an intervening lookup, lookup by id on the hard link
* item. This is only an issue for MAC hooks which cannot reenter the filesystem
* which currently are the MAC hooks for rename, unlink and rmdir.
*/
#define MAX_AUTHORIZE_ENOENT_RETRIES 1024
/* Max retry limit for rename due to vnode recycling. */
#define MAX_RENAME_ERECYCLE_RETRIES 1024
static int rmdirat_internal(vfs_context_t, int, user_addr_t, enum uio_seg,
int unlink_flags);
#ifdef CONFIG_IMGSRC_ACCESS
static int authorize_devpath_and_update_mntfromname(mount_t mp, user_addr_t devpath, vnode_t *devvpp, vfs_context_t ctx);
static int place_mount_and_checkdirs(mount_t mp, vnode_t vp, vfs_context_t ctx);
static void undo_place_on_covered_vp(mount_t mp, vnode_t vp);
static int mount_begin_update(mount_t mp, vfs_context_t ctx, int flags);
static void mount_end_update(mount_t mp);
static int relocate_imageboot_source(vnode_t pvp, vnode_t vp, struct componentname *cnp, const char *fsname, vfs_context_t ctx, boolean_t is64bit, user_addr_t fsmountargs, boolean_t by_index);
#endif /* CONFIG_IMGSRC_ACCESS */
//snapshot functions
#if CONFIG_MNT_ROOTSNAP
static int __attribute__ ((noinline)) snapshot_root(int dirfd, user_addr_t name, uint32_t flags, vfs_context_t ctx);
#else
static int __attribute__ ((noinline)) snapshot_root(int dirfd, user_addr_t name, uint32_t flags, vfs_context_t ctx) __attribute__((unused));
#endif
__private_extern__
int sync_internal(void);
__private_extern__
int unlink1(vfs_context_t, vnode_t, user_addr_t, enum uio_seg, int);
static LCK_GRP_DECLARE(fd_vn_lck_grp, "fd_vnode_data");
static LCK_ATTR_DECLARE(fd_vn_lck_attr, 0, 0);
/* vars for sync mutex */
static LCK_GRP_DECLARE(sync_mtx_lck_grp, "sync thread");
static LCK_MTX_DECLARE(sync_mtx_lck, &sync_mtx_lck_grp);
extern lck_rw_t rootvnode_rw_lock;
VFS_SMR_DECLARE;
extern uint32_t nc_smr_enabled;
/*
* incremented each time a mount or unmount operation occurs
* used to invalidate the cached value of the rootvp in the
* mount structure utilized by cache_lookup_path
*/
uint32_t mount_generation = 0;
/* counts number of mount and unmount operations */
unsigned int vfs_nummntops = 0;
/* system-wide, per-boot unique mount ID */
static _Atomic uint64_t mount_unique_id = 1;
extern const struct fileops vnops;
#if CONFIG_APPLEDOUBLE
extern errno_t rmdir_remove_orphaned_appleDouble(vnode_t, vfs_context_t, int *);
#endif /* CONFIG_APPLEDOUBLE */
/* Maximum buffer length supported by fsgetpath(2) */
#define FSGETPATH_MAXBUFLEN 8192
/*
* Virtual File System System Calls
*/
/*
* Private in-kernel mounting spi (specific use-cases only)
*/
boolean_t
vfs_iskernelmount(mount_t mp)
{
return (mp->mnt_kern_flag & MNTK_KERNEL_MOUNT) ? TRUE : FALSE;
}
__private_extern__
int
kernel_mount(const char *fstype, vnode_t pvp, vnode_t vp, const char *path,
void *data, __unused size_t datalen, int syscall_flags, uint32_t kern_flags,
vfs_context_t ctx)
{
struct nameidata nd;
boolean_t did_namei;
int error;
NDINIT(&nd, LOOKUP, OP_MOUNT, FOLLOW | AUDITVNPATH1 | WANTPARENT,
UIO_SYSSPACE, CAST_USER_ADDR_T(path), ctx);
kern_flags &= KERNEL_MOUNT_SANITIZE_MASK;
/*
* Get the vnode to be covered if it's not supplied
*/
if (vp == NULLVP) {
error = namei(&nd);
if (error) {
if (kern_flags & (KERNEL_MOUNT_SNAPSHOT | KERNEL_MOUNT_VOLBYROLE_MASK)) {
printf("failed to locate mount-on path: %s ", path);
}
return error;
}
vp = nd.ni_vp;
pvp = nd.ni_dvp;
did_namei = TRUE;
} else {
char *pnbuf = CAST_DOWN(char *, path);
nd.ni_cnd.cn_pnbuf = pnbuf;
nd.ni_cnd.cn_pnlen = (int)(strlen(pnbuf) + 1);
did_namei = FALSE;
}
kern_flags |= KERNEL_MOUNT_KMOUNT;
error = mount_common(fstype, pvp, vp, &nd.ni_cnd, CAST_USER_ADDR_T(data),
syscall_flags, kern_flags, NULL, ctx);
if (did_namei) {
vnode_put(vp);
vnode_put(pvp);
nameidone(&nd);
}
return error;
}
int
vfs_mount_at_path(const char *fstype, const char *path,
vnode_t pvp, vnode_t vp, void *data, size_t datalen,
int mnt_flags, int flags)
{
int syscall_flags = MNT_AUTOMOUNTED | mnt_flags;
int error, km_flags = 0;
vfs_context_t ctx = (flags & VFS_MOUNT_FLAG_CURRENT_CONTEXT) ? vfs_context_current() : vfs_context_kernel();
/*
* This call is currently restricted to specific use cases.
*/
if ((strcmp(fstype, "lifs") != 0) && (strcmp(fstype, "nfs") != 0)) {
return ENOTSUP;
}
#if !defined(XNU_TARGET_OS_OSX)
if (strcmp(fstype, "lifs") == 0) {
syscall_flags |= MNT_NOEXEC;
}
#endif
if (flags & VFS_MOUNT_FLAG_NOAUTH) {
km_flags |= KERNEL_MOUNT_NOAUTH;
}
if (flags & VFS_MOUNT_FLAG_PERMIT_UNMOUNT) {
km_flags |= KERNEL_MOUNT_PERMIT_UNMOUNT;
}
error = kernel_mount(fstype, pvp, vp, path, data, datalen,
syscall_flags, km_flags, ctx);
if (error) {
printf("%s: mount on %s failed, error %d\n", __func__, path,
error);
}
return error;
}
/*
* Mount a file system.
*/
/* ARGSUSED */
int
mount(proc_t p, struct mount_args *uap, __unused int32_t *retval)
{
struct __mac_mount_args muap;
muap.type = uap->type;
muap.path = uap->path;
muap.flags = uap->flags;
muap.data = uap->data;
muap.mac_p = USER_ADDR_NULL;
return __mac_mount(p, &muap, retval);
}
int
fmount(__unused proc_t p, struct fmount_args *uap, __unused int32_t *retval)
{
struct componentname cn;
vfs_context_t ctx = vfs_context_current();
size_t dummy = 0;
int error;
int flags = uap->flags;
char fstypename[MFSNAMELEN];
char *labelstr = NULL; /* regular mount call always sets it to NULL for __mac_mount() */
vnode_t pvp;
vnode_t vp;
AUDIT_ARG(fd, uap->fd);
AUDIT_ARG(fflags, flags);
/* fstypename will get audited by mount_common */
/* Sanity check the flags */
if (flags & (MNT_IMGSRC_BY_INDEX | MNT_ROOTFS)) {
return ENOTSUP;
}
if (flags & MNT_UNION) {
return EPERM;
}
error = copyinstr(uap->type, fstypename, MFSNAMELEN, &dummy);
if (error) {
return error;
}
if ((error = file_vnode(uap->fd, &vp)) != 0) {
return error;
}
if ((error = vnode_getwithref(vp)) != 0) {
file_drop(uap->fd);
return error;
}
pvp = vnode_getparent(vp);
if (pvp == NULL) {
if (vp->v_mountedhere || (vp->v_flag & VROOT) != 0) {
error = EBUSY;
} else {
error = EINVAL;
}
vnode_put(vp);
file_drop(uap->fd);
return error;
}
memset(&cn, 0, sizeof(struct componentname));
cn.cn_pnbuf = zalloc_flags(ZV_NAMEI, Z_WAITOK);
cn.cn_pnlen = MAXPATHLEN;
if ((error = vn_getpath(vp, cn.cn_pnbuf, &cn.cn_pnlen)) != 0) {
zfree(ZV_NAMEI, cn.cn_pnbuf);
vnode_put(pvp);
vnode_put(vp);
file_drop(uap->fd);
return error;
}
error = mount_common(fstypename, pvp, vp, &cn, uap->data, flags, KERNEL_MOUNT_FMOUNT, labelstr, ctx);
zfree(ZV_NAMEI, cn.cn_pnbuf);
vnode_put(pvp);
vnode_put(vp);
file_drop(uap->fd);
return error;
}
#define MAX_GRAFT_METADATA_SIZE 16384 /* bytes */
/*
* Get the size of a graft file (a manifest or payload file).
* The vp should be an iocounted vnode.
*/
static int
get_and_verify_graft_metadata_vp_size(vnode_t graft_vp, vfs_context_t vctx, size_t *size)
{
struct stat64 sb = {};
int error;
*size = 0;
error = vn_stat(graft_vp, &sb, NULL, 1, 0, vctx);
if (error) {
return error;
}
if (sb.st_size == 0) {
error = ENODATA;
} else if ((size_t) sb.st_size > MAX_GRAFT_METADATA_SIZE) {
error = EFBIG;
} else {
*size = (size_t) sb.st_size;
}
return error;
}
/*
* Read in a graft file (a manifest or payload file) of size `size` into `buf`.
* `size` must already be validated.
*/
static int
read_graft_metadata_vp(vnode_t graft_vp, vfs_context_t vctx, size_t size, void *buf)
{
return vn_rdwr(UIO_READ, graft_vp,
(caddr_t) buf, (int) size, /* offset */ 0,
UIO_SYSSPACE, IO_NOCACHE | IO_RAOFF | IO_UNIT,
vfs_context_ucred(vctx), /* resid */ NULL,
vfs_context_proc(vctx));
}
/*
* Convert a single graft file descriptor into a vnode, get its size (saving it to `size`),
* and read it into `buf`.
*/
static int
graft_secureboot_read_fd(int fd, vfs_context_t vctx, size_t *size, void *buf)
{
vnode_t metadata_vp = NULLVP;
int error;
// Convert this graft fd to a vnode.
if ((error = vnode_getfromfd(vctx, fd, &metadata_vp)) != 0) {
goto out;
}
// Get (and validate) size information.
if ((error = get_and_verify_graft_metadata_vp_size(metadata_vp, vctx, size)) != 0) {
goto out;
}
// Read each file into the provided buffer - we must get the expected amount of bytes.
if ((error = read_graft_metadata_vp(metadata_vp, vctx, *size, buf)) != 0) {
goto out;
}
out:
if (metadata_vp) {
vnode_put(metadata_vp);
metadata_vp = NULLVP;
}
return error;
}
/*
* Read graft file descriptors into buffers of size MAX_GRAFT_METADATA_SIZE
* provided in `gfs`, saving the size of data read in `gfs`.
*/
static int
graft_secureboot_read_metadata(secure_boot_cryptex_args_t *sbc_args, vfs_context_t vctx,
fsioc_graft_fs_t *gfs)
{
int error;
// Read the authentic manifest.
if ((error = graft_secureboot_read_fd(sbc_args->sbc_authentic_manifest_fd, vctx,
&gfs->authentic_manifest_size, gfs->authentic_manifest))) {
return error;
}
// The user manifest is currently unused, but set its size.
gfs->user_manifest_size = 0;
// Read the payload.
if ((error = graft_secureboot_read_fd(sbc_args->sbc_payload_fd, vctx,
&gfs->payload_size, gfs->payload))) {
return error;
}
return 0;
}
/*
* Call into the filesystem to verify and graft a cryptex.
*/
static int
graft_secureboot_cryptex(uint32_t graft_type, secure_boot_cryptex_args_t *sbc_args,
vfs_context_t vctx, vnode_t cryptex_vp, vnode_t mounton_vp)
{
fsioc_graft_fs_t gfs = {};
uint64_t graft_dir_ino = 0;
struct stat64 sb = {};
int error;
// Pre-flight arguments.
if (sbc_args->sbc_version != GRAFTDMG_SECURE_BOOT_CRYPTEX_ARGS_VERSION) {
// Make sure that this graft version matches what we support.
return ENOTSUP;
} else if (mounton_vp && cryptex_vp->v_mount != mounton_vp->v_mount) {
// For this type, cryptex VP must live on same volume as the target of graft.
return EXDEV;
} else if (mounton_vp && mounton_vp->v_type != VDIR) {
// We cannot graft upon non-directories.
return ENOTDIR;
} else if (sbc_args->sbc_authentic_manifest_fd < 0 ||
sbc_args->sbc_payload_fd < 0) {
// We cannot graft without a manifest and payload.
return EINVAL;
}
if (mounton_vp) {
// Get the mounton's inode number.
error = vn_stat(mounton_vp, &sb, NULL, 1, 0, vctx);
if (error) {
return error;
}
graft_dir_ino = (uint64_t) sb.st_ino;
}
// Create buffers (of our maximum-defined size) to store authentication info.
gfs.authentic_manifest = kalloc_data(MAX_GRAFT_METADATA_SIZE, Z_WAITOK | Z_ZERO);
gfs.payload = kalloc_data(MAX_GRAFT_METADATA_SIZE, Z_WAITOK | Z_ZERO);
if (!gfs.authentic_manifest || !gfs.payload) {
error = ENOMEM;
goto out;
}
// Read our fd's into our buffers.
// (Note that this will set the buffer size fields in `gfs`.)
error = graft_secureboot_read_metadata(sbc_args, vctx, &gfs);
if (error) {
goto out;
}
gfs.graft_version = FSIOC_GRAFT_VERSION;
gfs.graft_type = graft_type;
gfs.graft_4cc = sbc_args->sbc_4cc;
if (sbc_args->sbc_flags & SBC_PRESERVE_MOUNT) {
gfs.graft_flags |= FSCTL_GRAFT_PRESERVE_MOUNT;
}
if (sbc_args->sbc_flags & SBC_ALTERNATE_SHARED_REGION) {
gfs.graft_flags |= FSCTL_GRAFT_ALTERNATE_SHARED_REGION;
}
if (sbc_args->sbc_flags & SBC_SYSTEM_CONTENT) {
gfs.graft_flags |= FSCTL_GRAFT_SYSTEM_CONTENT;
}
if (sbc_args->sbc_flags & SBC_PANIC_ON_AUTHFAIL) {
gfs.graft_flags |= FSCTL_GRAFT_PANIC_ON_AUTHFAIL;
}
if (sbc_args->sbc_flags & SBC_STRICT_AUTH) {
gfs.graft_flags |= FSCTL_GRAFT_STRICT_AUTH;
}
if (sbc_args->sbc_flags & SBC_PRESERVE_GRAFT) {
gfs.graft_flags |= FSCTL_GRAFT_PRESERVE_GRAFT;
}
gfs.dir_ino = graft_dir_ino; // ino from mounton_vp (if not provided, the parent directory)
// Call into the FS to perform the graft (and validation).
error = VNOP_IOCTL(cryptex_vp, FSIOC_GRAFT_FS, (caddr_t)&gfs, 0, vctx);
out:
if (gfs.authentic_manifest) {
kfree_data(gfs.authentic_manifest, MAX_GRAFT_METADATA_SIZE);
gfs.authentic_manifest = NULL;
}
if (gfs.payload) {
kfree_data(gfs.payload, MAX_GRAFT_METADATA_SIZE);
gfs.payload = NULL;
}
return error;
}
#define GRAFTDMG_ENTITLEMENT "com.apple.private.vfs.graftdmg"
/*
* Graft a cryptex disk image (via FD) onto the appropriate mount-point
* { int graftdmg(int dmg_fd, const char *mountdir, uint32_t graft_type, graftdmg_args_un *gda); }
*/
int
graftdmg(__unused proc_t p, struct graftdmg_args *uap, __unused int32_t *retval)
{
int ua_dmgfd = uap->dmg_fd;
user_addr_t ua_mountdir = uap->mountdir;
uint32_t ua_grafttype = uap->graft_type;
user_addr_t ua_graftargs = uap->gda;
graftdmg_args_un kern_gda = {};
int error = 0;
secure_boot_cryptex_args_t *sbc_args = NULL;
vnode_t cryptex_vp = NULLVP;
vnode_t mounton_vp = NULLVP;
struct nameidata nd = {};
vfs_context_t ctx = vfs_context_current();
if (!IOTaskHasEntitlement(vfs_context_task(ctx), GRAFTDMG_ENTITLEMENT)) {
return EPERM;
}
error = copyin(ua_graftargs, &kern_gda, sizeof(graftdmg_args_un));
if (error) {
return error;
}
// Copy mount dir in, if provided.
if (ua_mountdir != USER_ADDR_NULL) {
// Acquire vnode for mount-on path
NDINIT(&nd, LOOKUP, OP_MOUNT, (FOLLOW | AUDITVNPATH1),
UIO_USERSPACE, ua_mountdir, ctx);
error = namei(&nd);
if (error) {
return error;
}
mounton_vp = nd.ni_vp;
}
// Convert fd to vnode.
error = vnode_getfromfd(ctx, ua_dmgfd, &cryptex_vp);
if (error) {
goto graftout;
}
if (ua_grafttype == 0 || ua_grafttype > GRAFTDMG_CRYPTEX_MAX) {
error = EINVAL;
} else {
sbc_args = &kern_gda.sbc_args;
error = graft_secureboot_cryptex(ua_grafttype, sbc_args, ctx, cryptex_vp, mounton_vp);
}
graftout:
if (cryptex_vp) {
vnode_put(cryptex_vp);
cryptex_vp = NULLVP;
}
if (mounton_vp) {
vnode_put(mounton_vp);
mounton_vp = NULLVP;
}
if (ua_mountdir != USER_ADDR_NULL) {
nameidone(&nd);
}
return error;
}
/*
* Ungraft a cryptex disk image (via mount dir FD)
* { int ungraftdmg(const char *mountdir, uint64_t flags); }
*/
int
ungraftdmg(__unused proc_t p, struct ungraftdmg_args *uap, __unused int32_t *retval)
{
int error = 0;
user_addr_t ua_mountdir = uap->mountdir;
fsioc_ungraft_fs_t ugfs;
vnode_t mounton_vp = NULLVP;
struct nameidata nd = {};
vfs_context_t ctx = vfs_context_current();
if (!IOTaskHasEntitlement(vfs_context_task(ctx), GRAFTDMG_ENTITLEMENT)) {
return EPERM;
}
if (uap->flags != 0 || ua_mountdir == USER_ADDR_NULL) {
return EINVAL;
}
ugfs.ungraft_flags = 0;
// Acquire vnode for mount-on path
NDINIT(&nd, LOOKUP, OP_MOUNT, (FOLLOW | AUDITVNPATH1),
UIO_USERSPACE, ua_mountdir, ctx);
error = namei(&nd);
if (error) {
return error;
}
mounton_vp = nd.ni_vp;
// Call into the FS to perform the ungraft
error = VNOP_IOCTL(mounton_vp, FSIOC_UNGRAFT_FS, (caddr_t)&ugfs, 0, ctx);
vnode_put(mounton_vp);
nameidone(&nd);
return error;
}
void
vfs_notify_mount(vnode_t pdvp)
{
vfs_event_signal(NULL, VQ_MOUNT, (intptr_t)NULL);
lock_vnode_and_post(pdvp, NOTE_WRITE);
}
/*
* __mac_mount:
* Mount a file system taking into account MAC label behavior.
* See mount(2) man page for more information
*
* Parameters: p Process requesting the mount
* uap User argument descriptor (see below)
* retval (ignored)
*
* Indirect: uap->type Filesystem type
* uap->path Path to mount
* uap->data Mount arguments
* uap->mac_p MAC info
* uap->flags Mount flags
*
*
* Returns: 0 Success
* !0 Not success
*/
boolean_t root_fs_upgrade_try = FALSE;
int
__mac_mount(struct proc *p, register struct __mac_mount_args *uap, __unused int32_t *retval)
{
vnode_t pvp = NULL;
vnode_t vp = NULL;
int need_nameidone = 0;
vfs_context_t ctx = vfs_context_current();
char fstypename[MFSNAMELEN];
struct nameidata nd;
size_t dummy = 0;
char *labelstr = NULL;
size_t labelsz = 0;
int flags = uap->flags;
int error;
#if CONFIG_IMGSRC_ACCESS || CONFIG_MACF
boolean_t is_64bit = IS_64BIT_PROCESS(p);
#else
#pragma unused(p)
#endif
/*
* Get the fs type name from user space
*/
error = copyinstr(uap->type, fstypename, MFSNAMELEN, &dummy);
if (error) {
return error;
}
/*
* Get the vnode to be covered
*/
NDINIT(&nd, LOOKUP, OP_MOUNT, FOLLOW | AUDITVNPATH1 | WANTPARENT,
UIO_USERSPACE, uap->path, ctx);
if (flags & MNT_NOFOLLOW) {
nd.ni_flag |= NAMEI_NOFOLLOW_ANY;
}
error = namei(&nd);
if (error) {
goto out;
}
need_nameidone = 1;
vp = nd.ni_vp;
pvp = nd.ni_dvp;
#ifdef CONFIG_IMGSRC_ACCESS
/* Mounting image source cannot be batched with other operations */
if (flags == MNT_IMGSRC_BY_INDEX) {
error = relocate_imageboot_source(pvp, vp, &nd.ni_cnd, fstypename,
ctx, is_64bit, uap->data, (flags == MNT_IMGSRC_BY_INDEX));
goto out;
}
#endif /* CONFIG_IMGSRC_ACCESS */
#if CONFIG_MACF
/*
* Get the label string (if any) from user space
*/
if (uap->mac_p != USER_ADDR_NULL) {
struct user_mac mac;
size_t ulen = 0;
if (is_64bit) {
struct user64_mac mac64;
error = copyin(uap->mac_p, &mac64, sizeof(mac64));
mac.m_buflen = (user_size_t)mac64.m_buflen;
mac.m_string = (user_addr_t)mac64.m_string;
} else {
struct user32_mac mac32;
error = copyin(uap->mac_p, &mac32, sizeof(mac32));
mac.m_buflen = mac32.m_buflen;
mac.m_string = mac32.m_string;
}
if (error) {
goto out;
}
if ((mac.m_buflen > MAC_MAX_LABEL_BUF_LEN) ||
(mac.m_buflen < 2)) {
error = EINVAL;
goto out;
}
labelsz = mac.m_buflen;
labelstr = kalloc_data(labelsz, Z_WAITOK);
error = copyinstr(mac.m_string, labelstr, mac.m_buflen, &ulen);
if (error) {
goto out;
}
AUDIT_ARG(mac_string, labelstr);
}
#endif /* CONFIG_MACF */
AUDIT_ARG(fflags, flags);
#if !CONFIG_UNION_MOUNTS
if (flags & MNT_UNION) {
error = EPERM;
goto out;
}
#endif
if ((vp->v_flag & VROOT) &&
(vp->v_mount->mnt_flag & MNT_ROOTFS)) {
#if CONFIG_UNION_MOUNTS
if (!(flags & MNT_UNION)) {
flags |= MNT_UPDATE;
} else {
/*
* For a union mount on '/', treat it as fresh
* mount instead of update.
* Otherwise, union mouting on '/' used to panic the
* system before, since mnt_vnodecovered was found to
* be NULL for '/' which is required for unionlookup
* after it gets ENOENT on union mount.
*/
flags = (flags & ~(MNT_UPDATE));
}
#else
flags |= MNT_UPDATE;
#endif /* CONFIG_UNION_MOUNTS */
#if SECURE_KERNEL
if ((flags & MNT_RDONLY) == 0) {
/* Release kernels are not allowed to mount "/" as rw */
error = EPERM;
goto out;
}
#endif
/*
* See 7392553 for more details on why this check exists.
* Suffice to say: If this check is ON and something tries
* to mount the rootFS RW, we'll turn off the codesign
* bitmap optimization.
*/
#if CHECK_CS_VALIDATION_BITMAP
if ((flags & MNT_RDONLY) == 0) {
root_fs_upgrade_try = TRUE;
}
#endif
}
error = mount_common(fstypename, pvp, vp, &nd.ni_cnd, uap->data, flags, 0,
labelstr, ctx);
out:
#if CONFIG_MACF
kfree_data(labelstr, labelsz);
#endif /* CONFIG_MACF */
if (vp) {
vnode_put(vp);
}
if (pvp) {
vnode_put(pvp);
}
if (need_nameidone) {
nameidone(&nd);
}
return error;
}
/*
* common mount implementation (final stage of mounting)
*
* Arguments:
* fstypename file system type (ie it's vfs name)
* pvp parent of covered vnode
* vp covered vnode
* cnp component name (ie path) of covered vnode
* flags generic mount flags
* fsmountargs file system specific data
* labelstr optional MAC label
* kernelmount TRUE for mounts initiated from inside the kernel
* ctx caller's context
*/
static int
mount_common(const char *fstypename, vnode_t pvp, vnode_t vp,
struct componentname *cnp, user_addr_t fsmountargs, int flags, uint32_t internal_flags,
char *labelstr, vfs_context_t ctx)
{
#if !CONFIG_MACF
#pragma unused(labelstr)
#endif
struct vnode *devvp = NULLVP;
struct vnode *device_vnode = NULLVP;
#if CONFIG_MACF
struct vnode *rvp;
#endif
struct mount *mp = NULL;
struct vfstable *vfsp = (struct vfstable *)0;
struct proc *p = vfs_context_proc(ctx);
int error, flag = 0;
bool flag_set = false;
user_addr_t devpath = USER_ADDR_NULL;
int ronly = 0;
int mntalloc = 0;
boolean_t vfsp_ref = FALSE;
boolean_t is_rwlock_locked = FALSE;
boolean_t did_rele = FALSE;
boolean_t have_usecount = FALSE;
boolean_t did_set_lmount = FALSE;
boolean_t kernelmount = !!(internal_flags & KERNEL_MOUNT_KMOUNT);
#if CONFIG_ROSV_STARTUP || CONFIG_MOUNT_VM || CONFIG_BASESYSTEMROOT
/* Check for mutually-exclusive flag bits */
uint32_t checkflags = (internal_flags & (KERNEL_MOUNT_VOLBYROLE_MASK | KERNEL_MOUNT_BASESYSTEMROOT));
int bitcount = 0;
while (checkflags != 0) {
checkflags &= (checkflags - 1);
bitcount++;
}
if (bitcount > 1) {
//not allowed to request multiple mount-by-role flags
error = EINVAL;
goto out1;
}
#endif
/*
* Process an update for an existing mount
*/
if (flags & MNT_UPDATE) {
if ((vp->v_flag & VROOT) == 0) {
error = EINVAL;
goto out1;
}
mp = vp->v_mount;
/* if unmount or mount in progress, return error */
mount_lock_spin(mp);
if (mp->mnt_lflag & (MNT_LUNMOUNT | MNT_LMOUNT)) {
mount_unlock(mp);
error = EBUSY;
goto out1;
}
mp->mnt_lflag |= MNT_LMOUNT;
did_set_lmount = TRUE;
mount_unlock(mp);
lck_rw_lock_exclusive(&mp->mnt_rwlock);
is_rwlock_locked = TRUE;
/*
* We only allow the filesystem to be reloaded if it
* is currently mounted read-only.
*/
if ((flags & MNT_RELOAD) &&
((mp->mnt_flag & MNT_RDONLY) == 0)) {
error = ENOTSUP;
goto out1;
}
/*
* If content protection is enabled, update mounts are not
* allowed to turn it off.
*/
if ((mp->mnt_flag & MNT_CPROTECT) &&
((flags & MNT_CPROTECT) == 0)) {
error = EINVAL;
goto out1;
}
/*
* can't turn off MNT_REMOVABLE either but it may be an unexpected
* failure to return an error for this so we'll just silently
* add it if it is not passed in.
*/
if ((mp->mnt_flag & MNT_REMOVABLE) &&
((flags & MNT_REMOVABLE) == 0)) {
flags |= MNT_REMOVABLE;
}
/* Can't downgrade the backer of the root FS */
if ((mp->mnt_kern_flag & MNTK_BACKS_ROOT) &&
(!vfs_isrdonly(mp)) && (flags & MNT_RDONLY)) {
error = ENOTSUP;
goto out1;
}
/*
* Only root, or the user that did the original mount is
* permitted to update it.
*/
if (mp->mnt_vfsstat.f_owner != kauth_cred_getuid(vfs_context_ucred(ctx)) &&
(error = suser(vfs_context_ucred(ctx), &p->p_acflag))) {
goto out1;
}
#if CONFIG_MACF
error = mac_mount_check_remount(ctx, mp);
if (error != 0) {
goto out1;
}
#endif
/*
* For non-root users, silently enforce MNT_NOSUID and MNT_NODEV,
* and MNT_NOEXEC if mount point is already MNT_NOEXEC.
*/
if ((!kernelmount) && suser(vfs_context_ucred(ctx), NULL)) {
flags |= MNT_NOSUID | MNT_NODEV;
if (mp->mnt_flag & MNT_NOEXEC) {
flags |= MNT_NOEXEC;
}
}
flag = mp->mnt_flag;
flag_set = true;
mp->mnt_flag |= flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE);
vfsp = mp->mnt_vtable;
goto update;
} // MNT_UPDATE
/*
* For non-root users, silently enforce MNT_NOSUID and MNT_NODEV, and
* MNT_NOEXEC if mount point is already MNT_NOEXEC.
*/
if ((!kernelmount) && suser(vfs_context_ucred(ctx), NULL)) {
flags |= MNT_NOSUID | MNT_NODEV;
if (vp->v_mount->mnt_flag & MNT_NOEXEC) {
flags |= MNT_NOEXEC;
}
}
/* XXXAUDIT: Should we capture the type on the error path as well? */
/* XXX cast-away const (audit_arg_text() does not modify its input) */
AUDIT_ARG(text, (char *)(uintptr_t)fstypename);
mount_list_lock();
for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) {
if (!strncmp(vfsp->vfc_name, fstypename, MFSNAMELEN)) {
vfsp->vfc_refcount++;
vfsp_ref = TRUE;
break;
}
}
mount_list_unlock();
if (vfsp == NULL) {
error = ENODEV;
goto out1;
}
/*
* VFC_VFSLOCALARGS is not currently supported for kernel mounts,
* except in ROSV configs and for the initial BaseSystem root.
*/
if (kernelmount && (vfsp->vfc_vfsflags & VFC_VFSLOCALARGS) &&
((internal_flags & KERNEL_MOUNT_VOLBYROLE_MASK) == 0) &&
((internal_flags & KERNEL_MOUNT_BASESYSTEMROOT) == 0)) {
error = EINVAL; /* unsupported request */
goto out1;
}
error = prepare_coveredvp(vp, ctx, cnp, fstypename, internal_flags);
if (error != 0) {
goto out1;
}
/*
* Allocate and initialize the filesystem (mount_t)
*/
mp = zalloc_flags(mount_zone, Z_WAITOK | Z_ZERO);
mntalloc = 1;
/* Initialize the default IO constraints */
mp->mnt_maxreadcnt = mp->mnt_maxwritecnt = MAXPHYS;
mp->mnt_segreadcnt = mp->mnt_segwritecnt = 32;
mp->mnt_maxsegreadsize = mp->mnt_maxreadcnt;
mp->mnt_maxsegwritesize = mp->mnt_maxwritecnt;
mp->mnt_devblocksize = DEV_BSIZE;
mp->mnt_alignmentmask = PAGE_MASK;
mp->mnt_ioqueue_depth = MNT_DEFAULT_IOQUEUE_DEPTH;
mp->mnt_ioscale = 1;
mp->mnt_ioflags = 0;
mp->mnt_realrootvp = NULLVP;
mp->mnt_authcache_ttl = CACHED_LOOKUP_RIGHT_TTL;
mp->mnt_lflag |= MNT_LMOUNT;
did_set_lmount = TRUE;
TAILQ_INIT(&mp->mnt_vnodelist);
TAILQ_INIT(&mp->mnt_workerqueue);
TAILQ_INIT(&mp->mnt_newvnodes);
mount_lock_init(mp);
lck_rw_lock_exclusive(&mp->mnt_rwlock);
is_rwlock_locked = TRUE;
mp->mnt_op = vfsp->vfc_vfsops;
mp->mnt_vtable = vfsp;
//mp->mnt_stat.f_type = vfsp->vfc_typenum;
mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK;
strlcpy(mp->mnt_vfsstat.f_fstypename, vfsp->vfc_name, MFSTYPENAMELEN);
do {
size_t pathlen = MAXPATHLEN;
if (vn_getpath_ext(vp, pvp, mp->mnt_vfsstat.f_mntonname, &pathlen, VN_GETPATH_FSENTER)) {
strlcpy(mp->mnt_vfsstat.f_mntonname, cnp->cn_pnbuf, MAXPATHLEN);
}
} while (0);
mp->mnt_vnodecovered = vp;
mp->mnt_vfsstat.f_owner = kauth_cred_getuid(vfs_context_ucred(ctx));
mp->mnt_throttle_mask = LOWPRI_MAX_NUM_DEV - 1;
mp->mnt_devbsdunit = 0;
mp->mnt_mount_id = os_atomic_inc_orig(&mount_unique_id, relaxed);
/* XXX 3762912 hack to support HFS filesystem 'owner' - filesystem may update later */
vfs_setowner(mp, KAUTH_UID_NONE, KAUTH_GID_NONE);
if (kernelmount) {
mp->mnt_kern_flag |= MNTK_KERNEL_MOUNT;
}
if ((internal_flags & KERNEL_MOUNT_PERMIT_UNMOUNT) != 0) {
mp->mnt_kern_flag |= MNTK_PERMIT_UNMOUNT;
}
if (KERNEL_MOUNT_DEVFS & internal_flags) {
// kernel mounted devfs
mp->mnt_kern_flag |= MNTK_SYSTEM;
}
update:
/*
* Set the mount level flags.
*/
if (flags & MNT_RDONLY) {
mp->mnt_flag |= MNT_RDONLY;
} else if (mp->mnt_flag & MNT_RDONLY) {
// disallow read/write upgrades of file systems that
// had the TYPENAME_OVERRIDE feature set.
if (mp->mnt_kern_flag & MNTK_TYPENAME_OVERRIDE) {
error = EPERM;
goto out1;
}
mp->mnt_kern_flag |= MNTK_WANTRDWR;
}
mp->mnt_flag &= ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC |
MNT_UNKNOWNPERMISSIONS | MNT_DONTBROWSE |
MNT_AUTOMOUNTED | MNT_DEFWRITE | MNT_NOATIME | MNT_STRICTATIME |
MNT_QUARANTINE | MNT_CPROTECT);
#if SECURE_KERNEL
#if !CONFIG_MNT_SUID
/*
* On release builds of iOS based platforms, always enforce NOSUID on
* all mounts. We do this here because we can catch update mounts as well as
* non-update mounts in this case.
*/
mp->mnt_flag |= (MNT_NOSUID);
#endif
#endif
mp->mnt_flag |= flags & (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC |
MNT_UNKNOWNPERMISSIONS | MNT_DONTBROWSE |
MNT_AUTOMOUNTED | MNT_DEFWRITE | MNT_NOATIME | MNT_STRICTATIME |
MNT_QUARANTINE | MNT_CPROTECT);
#if CONFIG_MACF
if (flags & MNT_MULTILABEL) {
if (vfsp->vfc_vfsflags & VFC_VFSNOMACLABEL) {
error = EINVAL;
goto out1;
}
mp->mnt_flag |= MNT_MULTILABEL;
}
#endif
/*
* Process device path for local file systems if requested.
*
* Snapshot and mount-by-role mounts do not use this path; they are
* passing other opaque data in the device path field.
*
* Basesystemroot mounts pass a device path to be resolved here,
* but it's just a char * already inside the kernel, which
* kernel_mount() shoved into a user_addr_t to call us. So for such
* mounts we must skip copyin (both of the address and of the string
* (in NDINIT).
*/
if (vfsp->vfc_vfsflags & VFC_VFSLOCALARGS &&
!(internal_flags & (KERNEL_MOUNT_SNAPSHOT | KERNEL_MOUNT_VOLBYROLE_MASK))) {
boolean_t do_copyin_devpath = true;
#if CONFIG_BASESYSTEMROOT
if (internal_flags & KERNEL_MOUNT_BASESYSTEMROOT) {
// KERNEL_MOUNT_BASESYSTEMROOT implies subtle behavior worh nothing:
// We have been passed fsmountargs, which is typed as a user_addr_t,
// but is actually a char ** pointing to a (kernelspace) string.
// We manually unpack it with a series of casts and dereferences
// that reverses what was done just above us on the stack in
// imageboot_pivot_image().
// After retrieving the path to the dev node (which we will NDINIT
// in a moment), we pass NULL fsmountargs on to the filesystem.
_Static_assert(sizeof(char **) == sizeof(fsmountargs), "fsmountargs should fit a (kernel) address");
char **devnamepp = (char **)fsmountargs;
char *devnamep = *devnamepp;
devpath = CAST_USER_ADDR_T(devnamep);
do_copyin_devpath = false;
fsmountargs = USER_ADDR_NULL;
//Now that we have a mp, denote that this mount is for the basesystem.
mp->mnt_supl_kern_flag |= MNTK_SUPL_BASESYSTEM;
}
#endif // CONFIG_BASESYSTEMROOT
if (do_copyin_devpath) {
if (vfs_context_is64bit(ctx)) {
if ((error = copyin(fsmountargs, (caddr_t)&devpath, sizeof(devpath)))) {
goto out1;
}
fsmountargs += sizeof(devpath);
} else {
user32_addr_t tmp;
if ((error = copyin(fsmountargs, (caddr_t)&tmp, sizeof(tmp)))) {
goto out1;
}
/* munge into LP64 addr */
devpath = CAST_USER_ADDR_T(tmp);
fsmountargs += sizeof(tmp);
}
}
/* Lookup device and authorize access to it */
if ((devpath)) {
struct nameidata nd;
enum uio_seg seg = UIO_USERSPACE;
#if CONFIG_BASESYSTEMROOT
if (internal_flags & KERNEL_MOUNT_BASESYSTEMROOT) {
seg = UIO_SYSSPACE;
}
#endif // CONFIG_BASESYSTEMROOT
NDINIT(&nd, LOOKUP, OP_MOUNT, FOLLOW, seg, devpath, ctx);
if ((error = namei(&nd))) {
goto out1;
}
strlcpy(mp->mnt_vfsstat.f_mntfromname, nd.ni_cnd.cn_pnbuf, MAXPATHLEN);
devvp = nd.ni_vp;
nameidone(&nd);
if (devvp->v_type != VBLK) {
error = ENOTBLK;
goto out2;
}
if (major(devvp->v_rdev) >= nblkdev) {
error = ENXIO;
goto out2;
}
/*
* If mount by non-root, then verify that user has necessary
* permissions on the device.
*/
if (suser(vfs_context_ucred(ctx), NULL) != 0) {
kauth_action_t accessmode = KAUTH_VNODE_READ_DATA;
if ((mp->mnt_flag & MNT_RDONLY) == 0) {
accessmode |= KAUTH_VNODE_WRITE_DATA;
}
if ((error = vnode_authorize(devvp, NULL, accessmode, ctx)) != 0) {
goto out2;
}
}
}
/* On first mount, preflight and open device */
if (devpath && ((flags & MNT_UPDATE) == 0)) {
if ((error = vnode_ref(devvp))) {
goto out2;
}
/*
* Disallow multiple mounts of the same device.
* Disallow mounting of a device that is currently in use
* (except for root, which might share swap device for miniroot).
* Flush out any old buffers remaining from a previous use.
*/
if ((error = vfs_setmounting(devvp))) {
vnode_rele(devvp);
goto out2;
}
if (vcount(devvp) > 1 && !(vfs_flags(mp) & MNT_ROOTFS)) {
error = EBUSY;
goto out3;
}
if ((error = VNOP_FSYNC(devvp, MNT_WAIT, ctx))) {
error = ENOTBLK;
goto out3;
}
if ((error = buf_invalidateblks(devvp, BUF_WRITE_DATA, 0, 0))) {
goto out3;
}
ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
#if CONFIG_MACF
error = mac_vnode_check_open(ctx,
devvp,
ronly ? FREAD : FREAD | FWRITE);
if (error) {
goto out3;
}
#endif /* MAC */
if ((error = VNOP_OPEN(devvp, ronly ? FREAD : FREAD | FWRITE, ctx))) {
goto out3;
}
mp->mnt_devvp = devvp;
device_vnode = devvp;
} else if ((mp->mnt_flag & MNT_RDONLY) &&
(mp->mnt_kern_flag & MNTK_WANTRDWR) &&
(device_vnode = mp->mnt_devvp)) {
dev_t dev;
int maj;
/*
* If upgrade to read-write by non-root, then verify
* that user has necessary permissions on the device.
*/
vnode_getalways(device_vnode);
if (suser(vfs_context_ucred(ctx), NULL) &&
(error = vnode_authorize(device_vnode, NULL,
KAUTH_VNODE_READ_DATA | KAUTH_VNODE_WRITE_DATA,
ctx)) != 0) {
vnode_put(device_vnode);
goto out2;
}
/* Tell the device that we're upgrading */
dev = (dev_t)device_vnode->v_rdev;
maj = major(dev);
if ((u_int)maj >= (u_int)nblkdev) {
panic("Volume mounted on a device with invalid major number.");
}
error = bdevsw[maj].d_open(dev, FREAD | FWRITE, S_IFBLK, p);
vnode_put(device_vnode);
device_vnode = NULLVP;
if (error != 0) {
goto out2;
}
}
} // localargs && !(snapshot | data | vm)
#if CONFIG_MACF
if ((flags & MNT_UPDATE) == 0) {
mac_mount_label_init(mp);
mac_mount_label_associate(ctx, mp);
}
if (labelstr) {
if ((flags & MNT_UPDATE) != 0) {
error = mac_mount_check_label_update(ctx, mp);
if (error != 0) {
goto out3;
}
}
}
#endif
/*
* Mount the filesystem. We already asserted that internal_flags
* cannot have more than one mount-by-role bit set.
*/
if (internal_flags & KERNEL_MOUNT_SNAPSHOT) {
error = VFS_IOCTL(mp, VFSIOC_MOUNT_SNAPSHOT,
(caddr_t)fsmountargs, 0, ctx);
} else if (internal_flags & KERNEL_MOUNT_DATAVOL) {
#if CONFIG_ROSV_STARTUP
struct mount *origin_mp = (struct mount*)fsmountargs;
fs_role_mount_args_t frma = {origin_mp, VFS_DATA_ROLE};
error = VFS_IOCTL(mp, VFSIOC_MOUNT_BYROLE, (caddr_t)&frma, 0, ctx);
if (error) {
printf("MOUNT-BY-ROLE (%d) failed! (%d)", VFS_DATA_ROLE, error);
} else {
/* Mark volume associated with system volume */
mp->mnt_kern_flag |= MNTK_SYSTEM;
/* Attempt to acquire the mnt_devvp and set it up */
struct vnode *mp_devvp = NULL;
if (mp->mnt_vfsstat.f_mntfromname[0] != 0) {
errno_t lerr = vnode_lookup(mp->mnt_vfsstat.f_mntfromname,
0, &mp_devvp, vfs_context_kernel());
if (!lerr) {
mp->mnt_devvp = mp_devvp;
//vnode_lookup took an iocount, need to drop it.
vnode_put(mp_devvp);
// now set `device_vnode` to the devvp that was acquired.
// this is needed in order to ensure vfs_init_io_attributes is invoked.
// note that though the iocount above was dropped, the mount acquires
// an implicit reference against the device.
device_vnode = mp_devvp;
}
}
}
#else
error = EINVAL;
#endif
} else if (internal_flags & KERNEL_MOUNT_VMVOL) {
#if CONFIG_MOUNT_VM
struct mount *origin_mp = (struct mount*)fsmountargs;
fs_role_mount_args_t frma = {origin_mp, VFS_VM_ROLE};
error = VFS_IOCTL(mp, VFSIOC_MOUNT_BYROLE, (caddr_t)&frma, 0, ctx);
if (error) {
printf("MOUNT-BY-ROLE (%d) failed! (%d)", VFS_VM_ROLE, error);
} else {
/* Mark volume associated with system volume and a swap mount */
mp->mnt_kern_flag |= (MNTK_SYSTEM | MNTK_SWAP_MOUNT);
/* Attempt to acquire the mnt_devvp and set it up */
struct vnode *mp_devvp = NULL;
if (mp->mnt_vfsstat.f_mntfromname[0] != 0) {
errno_t lerr = vnode_lookup(mp->mnt_vfsstat.f_mntfromname,
0, &mp_devvp, vfs_context_kernel());
if (!lerr) {
mp->mnt_devvp = mp_devvp;
//vnode_lookup took an iocount, need to drop it.
vnode_put(mp_devvp);
// now set `device_vnode` to the devvp that was acquired.
// note that though the iocount above was dropped, the mount acquires
// an implicit reference against the device.
device_vnode = mp_devvp;
}
}
}
#else
error = EINVAL;
#endif
} else if ((internal_flags & KERNEL_MOUNT_PREBOOTVOL) || (internal_flags & KERNEL_MOUNT_RECOVERYVOL)) {
#if CONFIG_MOUNT_PREBOOTRECOVERY
struct mount *origin_mp = (struct mount*)fsmountargs;
uint32_t mount_role = 0;
if (internal_flags & KERNEL_MOUNT_PREBOOTVOL) {
mount_role = VFS_PREBOOT_ROLE;
} else if (internal_flags & KERNEL_MOUNT_RECOVERYVOL) {
mount_role = VFS_RECOVERY_ROLE;
}
if (mount_role != 0) {
fs_role_mount_args_t frma = {origin_mp, mount_role};
error = VFS_IOCTL(mp, VFSIOC_MOUNT_BYROLE, (caddr_t)&frma, 0, ctx);
if (error) {
printf("MOUNT-BY-ROLE (%d) failed! (%d)", mount_role, error);
} else {
// NOT YET - need to qualify how this interacts with shutdown, ERP/ERB, etc
/* Mark volume associated with system volume */
//mp->mnt_kern_flag |= MNTK_SYSTEM;
/* Attempt to acquire the mnt_devvp and set it up */
struct vnode *mp_devvp = NULL;
if (mp->mnt_vfsstat.f_mntfromname[0] != 0) {
errno_t lerr = vnode_lookup(mp->mnt_vfsstat.f_mntfromname,
0, &mp_devvp, vfs_context_kernel());
if (!lerr) {
mp->mnt_devvp = mp_devvp;
//vnode_lookup took an iocount, need to drop it.
vnode_put(mp_devvp);
// now set `device_vnode` to the devvp that was acquired.
// note that though the iocount above was dropped, the mount acquires
// an implicit reference against the device.
device_vnode = mp_devvp;
}
}
}
} else {
printf("MOUNT-BY-ROLE (%d) failed - ROLE UNRECOGNIZED! (%d)", mount_role, error);
error = EINVAL;
}
#else
error = EINVAL;
#endif
} else {
error = VFS_MOUNT(mp, device_vnode, fsmountargs, ctx);
}
if (flags & MNT_UPDATE) {
if (mp->mnt_kern_flag & MNTK_WANTRDWR) {
mp->mnt_flag &= ~MNT_RDONLY;
}
mp->mnt_flag &= ~
(MNT_UPDATE | MNT_RELOAD | MNT_FORCE);
mp->mnt_kern_flag &= ~MNTK_WANTRDWR;
if (error) {
mp->mnt_flag = flag; /* restore flag value */
}
vfs_event_signal(NULL, VQ_UPDATE, (intptr_t)NULL);
lck_rw_done(&mp->mnt_rwlock);
is_rwlock_locked = FALSE;
if (!error) {
enablequotas(mp, ctx);
}
goto exit;
}
/*
* Put the new filesystem on the mount list after root.
*/
if (error == 0) {
struct vfs_attr vfsattr;
if (device_vnode) {
/*
* cache the IO attributes for the underlying physical media...
* an error return indicates the underlying driver doesn't
* support all the queries necessary... however, reasonable
* defaults will have been set, so no reason to bail or care
*
* Need to do this before calling the MAC hook as it needs
* information from this call.
*/
vfs_init_io_attributes(device_vnode, mp);
}
#if CONFIG_MACF
error = mac_mount_check_mount_late(ctx, mp);
if (error != 0) {
goto out4;
}
if (vfs_flags(mp) & MNT_MULTILABEL) {
error = VFS_ROOT(mp, &rvp, ctx);
if (error) {
printf("%s() VFS_ROOT returned %d\n", __func__, error);
goto out4;
}
error = vnode_label(mp, NULL, rvp, NULL, 0, ctx);
/*
* drop reference provided by VFS_ROOT
*/
vnode_put(rvp);
if (error) {
goto out4;
}
}
#endif /* MAC */
vnode_lock_spin(vp);
CLR(vp->v_flag, VMOUNT);
vp->v_mountedhere = mp;
SET(vp->v_flag, VMOUNTEDHERE);
vnode_unlock(vp);
/*
* taking the name_cache_lock exclusively will
* insure that everyone is out of the fast path who
* might be trying to use a now stale copy of
* vp->v_mountedhere->mnt_realrootvp
* bumping mount_generation causes the cached values
* to be invalidated
*/
name_cache_lock();
mount_generation++;
name_cache_unlock();
error = vnode_ref(vp);
if (error != 0) {
goto out4;
}
have_usecount = TRUE;
error = checkdirs(vp, ctx);
if (error != 0) {
/* Unmount the filesystem as cdir/rdirs cannot be updated */
goto out4;
}
/*
* there is no cleanup code here so I have made it void
* we need to revisit this
*/
(void)VFS_START(mp, 0, ctx);
if (mount_list_add(mp) != 0) {
/*
* The system is shutting down trying to umount
* everything, so fail with a plausible errno.
*/
error = EBUSY;
goto out4;
}
lck_rw_done(&mp->mnt_rwlock);
is_rwlock_locked = FALSE;
/* Check if this mounted file system supports EAs or named streams. */
/* Skip WebDAV file systems for now since they hang in VFS_GETATTR here. */
VFSATTR_INIT(&vfsattr);
VFSATTR_WANTED(&vfsattr, f_capabilities);
if (strncmp(mp->mnt_vfsstat.f_fstypename, "webdav", sizeof("webdav")) != 0 &&
vfs_getattr(mp, &vfsattr, ctx) == 0 &&
VFSATTR_IS_SUPPORTED(&vfsattr, f_capabilities)) {
if ((vfsattr.f_capabilities.capabilities[VOL_CAPABILITIES_INTERFACES] & VOL_CAP_INT_EXTENDED_ATTR) &&
(vfsattr.f_capabilities.valid[VOL_CAPABILITIES_INTERFACES] & VOL_CAP_INT_EXTENDED_ATTR)) {
mp->mnt_kern_flag |= MNTK_EXTENDED_ATTRS;
}
#if NAMEDSTREAMS
if ((vfsattr.f_capabilities.capabilities[VOL_CAPABILITIES_INTERFACES] & VOL_CAP_INT_NAMEDSTREAMS) &&
(vfsattr.f_capabilities.valid[VOL_CAPABILITIES_INTERFACES] & VOL_CAP_INT_NAMEDSTREAMS)) {
mp->mnt_kern_flag |= MNTK_NAMED_STREAMS;
}
#endif
/* Check if this file system supports path from id lookups. */
if ((vfsattr.f_capabilities.capabilities[VOL_CAPABILITIES_FORMAT] & VOL_CAP_FMT_PATH_FROM_ID) &&
(vfsattr.f_capabilities.valid[VOL_CAPABILITIES_FORMAT] & VOL_CAP_FMT_PATH_FROM_ID)) {
mp->mnt_kern_flag |= MNTK_PATH_FROM_ID;
} else if (mp->mnt_flag & MNT_DOVOLFS) {
/* Legacy MNT_DOVOLFS flag also implies path from id lookups. */
mp->mnt_kern_flag |= MNTK_PATH_FROM_ID;
}
if ((vfsattr.f_capabilities.capabilities[VOL_CAPABILITIES_FORMAT] & VOL_CAP_FMT_DIR_HARDLINKS) &&
(vfsattr.f_capabilities.valid[VOL_CAPABILITIES_FORMAT] & VOL_CAP_FMT_DIR_HARDLINKS)) {
mp->mnt_kern_flag |= MNTK_DIR_HARDLINKS;
}
}
if (mp->mnt_vtable->vfc_vfsflags & VFC_VFSNATIVEXATTR) {
mp->mnt_kern_flag |= MNTK_EXTENDED_ATTRS;
}
if (mp->mnt_vtable->vfc_vfsflags & VFC_VFSPREFLIGHT) {
mp->mnt_kern_flag |= MNTK_UNMOUNT_PREFLIGHT;
}
/* increment the operations count */
OSAddAtomic(1, &vfs_nummntops);
enablequotas(mp, ctx);
if (device_vnode) {
vfs_setmountedon(device_vnode);
}
/* Now that mount is setup, notify the listeners */
vfs_notify_mount(pvp);
IOBSDMountChange(mp, kIOMountChangeMount);
} else {
/* If we fail a fresh mount, there should be no vnodes left hooked into the mountpoint. */
if (mp->mnt_vnodelist.tqh_first != NULL) {
panic("mount_common(): mount of %s filesystem failed with %d, but vnode list is not empty.",
mp->mnt_vtable->vfc_name, error);
}
vnode_lock_spin(vp);
CLR(vp->v_flag, VMOUNT);
vnode_unlock(vp);
mount_list_lock();
mp->mnt_vtable->vfc_refcount--;
mount_list_unlock();
if (device_vnode) {
vnode_rele(device_vnode);
VNOP_CLOSE(device_vnode, ronly ? FREAD : FREAD | FWRITE, ctx);
vfs_clearmounting(device_vnode);
}
lck_rw_done(&mp->mnt_rwlock);
is_rwlock_locked = FALSE;
if (nc_smr_enabled) {
vfs_smr_synchronize();
}
/*
* if we get here, we have a mount structure that needs to be freed,
* but since the coveredvp hasn't yet been updated to point at it,
* no need to worry about other threads holding a crossref on this mp
* so it's ok to just free it
*/
mount_lock_destroy(mp);
#if CONFIG_MACF
mac_mount_label_destroy(mp);
#endif
zfree(mount_zone, mp);
did_set_lmount = false;
}
exit:
/*
* drop I/O count on the device vp if there was one
*/
if (devpath && devvp) {
vnode_put(devvp);
}
if (did_set_lmount) {
mount_lock_spin(mp);
mp->mnt_lflag &= ~MNT_LMOUNT;
mount_unlock(mp);
}
return error;
/* Error condition exits */
out4:
(void)VFS_UNMOUNT(mp, MNT_FORCE, ctx);
/*
* If the mount has been placed on the covered vp,
* it may have been discovered by now, so we have
* to treat this just like an unmount
*/
mount_lock_spin(mp);
mp->mnt_lflag |= MNT_LDEAD;
mount_unlock(mp);
if (device_vnode != NULLVP) {
vnode_rele(device_vnode);
VNOP_CLOSE(device_vnode, mp->mnt_flag & MNT_RDONLY ? FREAD : FREAD | FWRITE,
ctx);
vfs_clearmounting(device_vnode);
did_rele = TRUE;
}
vnode_lock_spin(vp);
mp->mnt_crossref++;
CLR(vp->v_flag, VMOUNTEDHERE);
vp->v_mountedhere = (mount_t) 0;
vnode_unlock(vp);
if (have_usecount) {
vnode_rele(vp);
}
out3:
if (devpath && ((flags & MNT_UPDATE) == 0) && (!did_rele)) {
vnode_rele(devvp);
vfs_clearmounting(devvp);
}
out2:
if (devpath && devvp) {
vnode_put(devvp);
}
out1:
/* Release mnt_rwlock only when it was taken */
if (is_rwlock_locked == TRUE) {
if (flag_set) {
mp->mnt_flag = flag; /* restore mnt_flag value */
}
lck_rw_done(&mp->mnt_rwlock);
}
if (did_set_lmount) {
mount_lock_spin(mp);
mp->mnt_lflag &= ~MNT_LMOUNT;
mount_unlock(mp);
}
if (mntalloc) {
if (mp->mnt_crossref) {
mount_dropcrossref(mp, vp, 0);
} else {
if (nc_smr_enabled) {
vfs_smr_synchronize();
}
mount_lock_destroy(mp);
#if CONFIG_MACF
mac_mount_label_destroy(mp);
#endif
zfree(mount_zone, mp);
}
}
if (vfsp_ref) {
mount_list_lock();
vfsp->vfc_refcount--;
mount_list_unlock();
}
return error;
}
/*
* Flush in-core data, check for competing mount attempts,
* and set VMOUNT
*/
int
prepare_coveredvp(vnode_t vp, vfs_context_t ctx, struct componentname *cnp, const char *fsname, uint32_t internal_flags)
{
#if !CONFIG_MACF
#pragma unused(cnp,fsname)
#endif
struct vnode_attr va;
int error;
boolean_t skip_auth = !!(internal_flags & KERNEL_MOUNT_NOAUTH);
boolean_t is_fmount = !!(internal_flags & KERNEL_MOUNT_FMOUNT);
boolean_t is_busy;
if (!skip_auth) {
/*
* If the user is not root, ensure that they own the directory
* onto which we are attempting to mount.
*/
VATTR_INIT(&va);
VATTR_WANTED(&va, va_uid);
if ((error = vnode_getattr(vp, &va, ctx)) ||
(va.va_uid != kauth_cred_getuid(vfs_context_ucred(ctx)) &&
(!vfs_context_issuser(ctx)))) {
error = EPERM;
goto out;
}
}
if ((error = VNOP_FSYNC(vp, MNT_WAIT, ctx))) {
goto out;
}
if ((error = buf_invalidateblks(vp, BUF_WRITE_DATA, 0, 0))) {
goto out;
}
if (vp->v_type != VDIR) {
error = ENOTDIR;
goto out;
}
vnode_lock_spin(vp);
is_busy = is_fmount ?
(ISSET(vp->v_flag, VMOUNT) || (vp->v_mountedhere != NULL)) :
(ISSET(vp->v_flag, VMOUNT) && (vp->v_mountedhere != NULL));
if (is_busy) {
vnode_unlock(vp);
error = EBUSY;
goto out;
}
SET(vp->v_flag, VMOUNT);
vnode_unlock(vp);
#if CONFIG_MACF
error = mac_mount_check_mount(ctx, vp,
cnp, fsname);
if (error != 0) {
vnode_lock_spin(vp);
CLR(vp->v_flag, VMOUNT);
vnode_unlock(vp);
}
#endif
out:
return error;
}
#if CONFIG_IMGSRC_ACCESS
#define DEBUG_IMGSRC 0
#if DEBUG_IMGSRC
#define IMGSRC_DEBUG(args...) printf("imgsrc: " args)
#else
#define IMGSRC_DEBUG(args...) do { } while(0)
#endif
static int
authorize_devpath_and_update_mntfromname(mount_t mp, user_addr_t devpath, vnode_t *devvpp, vfs_context_t ctx)
{
struct nameidata nd;
vnode_t vp, realdevvp;
kauth_action_t accessmode;
int error;
enum uio_seg uio = UIO_USERSPACE;
if (ctx == vfs_context_kernel()) {
uio = UIO_SYSSPACE;
}
NDINIT(&nd, LOOKUP, OP_LOOKUP, FOLLOW, uio, devpath, ctx);
if ((error = namei(&nd))) {
IMGSRC_DEBUG("namei() failed with %d\n", error);
return error;
}
vp = nd.ni_vp;
if (!vnode_isblk(vp)) {
IMGSRC_DEBUG("Not block device.\n");
error = ENOTBLK;
goto out;
}
realdevvp = mp->mnt_devvp;
if (realdevvp == NULLVP) {
IMGSRC_DEBUG("No device backs the mount.\n");
error = ENXIO;
goto out;
}
error = vnode_getwithref(realdevvp);
if (error != 0) {
IMGSRC_DEBUG("Coudn't get iocount on device.\n");
goto out;
}
if (vnode_specrdev(vp) != vnode_specrdev(realdevvp)) {
IMGSRC_DEBUG("Wrong dev_t.\n");
error = ENXIO;
goto out1;
}
strlcpy(mp->mnt_vfsstat.f_mntfromname, nd.ni_cnd.cn_pnbuf, MAXPATHLEN);
/*
* If mount by non-root, then verify that user has necessary
* permissions on the device.
*/
if (!vfs_context_issuser(ctx)) {
accessmode = KAUTH_VNODE_READ_DATA;
if ((mp->mnt_flag & MNT_RDONLY) == 0) {
accessmode |= KAUTH_VNODE_WRITE_DATA;
}
if ((error = vnode_authorize(vp, NULL, accessmode, ctx)) != 0) {
IMGSRC_DEBUG("Access denied.\n");
goto out1;
}
}
*devvpp = vp;
out1:
vnode_put(realdevvp);
out:
nameidone(&nd);
if (error) {
vnode_put(vp);
}
return error;
}
/*
* Clear VMOUNT, set v_mountedhere, and mnt_vnodecovered, ref the vnode,
* and call checkdirs()
*/
static int
place_mount_and_checkdirs(mount_t mp, vnode_t vp, vfs_context_t ctx)
{
int error;
mp->mnt_vnodecovered = vp; /* XXX This is normally only set at init-time ... */
IMGSRC_DEBUG("placing: fsname = %s, vp = %s\n",
mp->mnt_vtable->vfc_name, vnode_getname(vp));
vnode_lock_spin(vp);
CLR(vp->v_flag, VMOUNT);
vp->v_mountedhere = mp;
SET(vp->v_flag, VMOUNTEDHERE);
vnode_unlock(vp);
/*
* taking the name_cache_lock exclusively will
* insure that everyone is out of the fast path who
* might be trying to use a now stale copy of
* vp->v_mountedhere->mnt_realrootvp
* bumping mount_generation causes the cached values
* to be invalidated
*/
name_cache_lock();
mount_generation++;
name_cache_unlock();
error = vnode_ref(vp);
if (error != 0) {
goto out;
}
error = checkdirs(vp, ctx);
if (error != 0) {
/* Unmount the filesystem as cdir/rdirs cannot be updated */
vnode_rele(vp);
goto out;
}
out:
if (error != 0) {
mp->mnt_vnodecovered = NULLVP;
}
return error;
}
static void
undo_place_on_covered_vp(mount_t mp, vnode_t vp)
{
vnode_rele(vp);
vnode_lock_spin(vp);
CLR(vp->v_flag, (VMOUNT | VMOUNTEDHERE));
vp->v_mountedhere = (mount_t)NULL;
vnode_unlock(vp);
mp->mnt_vnodecovered = NULLVP;
}
static int
mount_begin_update(mount_t mp, vfs_context_t ctx, int flags)
{
int error;
/* unmount in progress return error */
mount_lock_spin(mp);
if (mp->mnt_lflag & (MNT_LUNMOUNT | MNT_LMOUNT)) {
mount_unlock(mp);
return EBUSY;
}
mount_unlock(mp);
lck_rw_lock_exclusive(&mp->mnt_rwlock);
/*
* We only allow the filesystem to be reloaded if it
* is currently mounted read-only.
*/
if ((flags & MNT_RELOAD) &&
((mp->mnt_flag & MNT_RDONLY) == 0)) {
error = ENOTSUP;
goto out;
}
/*
* Only root, or the user that did the original mount is
* permitted to update it.
*/
if (mp->mnt_vfsstat.f_owner != kauth_cred_getuid(vfs_context_ucred(ctx)) &&
(!vfs_context_issuser(ctx))) {
error = EPERM;
goto out;
}
#if CONFIG_MACF
error = mac_mount_check_remount(ctx, mp);
if (error != 0) {
goto out;
}
#endif
out:
if (error) {
lck_rw_done(&mp->mnt_rwlock);
}
return error;
}
static void
mount_end_update(mount_t mp)
{
lck_rw_done(&mp->mnt_rwlock);
}
static int
get_imgsrc_rootvnode(uint32_t height, vnode_t *rvpp)
{
vnode_t vp;
if (height >= MAX_IMAGEBOOT_NESTING) {
return EINVAL;
}
vp = imgsrc_rootvnodes[height];
if ((vp != NULLVP) && (vnode_get(vp) == 0)) {
*rvpp = vp;
return 0;
} else {
return ENOENT;
}
}
static int
relocate_imageboot_source(vnode_t pvp, vnode_t vp,
struct componentname *cnp, const char *fsname, vfs_context_t ctx,
boolean_t is64bit, user_addr_t fsmountargs, boolean_t by_index)
{
int error;
mount_t mp;
boolean_t placed = FALSE;
struct vfstable *vfsp;
user_addr_t devpath;
char *old_mntonname;
vnode_t rvp;
vnode_t devvp;
uint32_t height;
uint32_t flags;
/* If we didn't imageboot, nothing to move */
if (imgsrc_rootvnodes[0] == NULLVP) {
return EINVAL;
}
/* Only root can do this */
if (!vfs_context_issuser(ctx)) {
return EPERM;
}
IMGSRC_DEBUG("looking for root vnode.\n");
/*
* Get root vnode of filesystem we're moving.
*/
if (by_index) {
if (is64bit) {
struct user64_mnt_imgsrc_args mia64;
error = copyin(fsmountargs, &mia64, sizeof(mia64));
if (error != 0) {
IMGSRC_DEBUG("Failed to copy in arguments.\n");
return error;
}
height = mia64.mi_height;
flags = mia64.mi_flags;
devpath = (user_addr_t)mia64.mi_devpath;
} else {
struct user32_mnt_imgsrc_args mia32;
error = copyin(fsmountargs, &mia32, sizeof(mia32));
if (error != 0) {
IMGSRC_DEBUG("Failed to copy in arguments.\n");
return error;
}
height = mia32.mi_height;
flags = mia32.mi_flags;
devpath = mia32.mi_devpath;
}
} else {
/*
* For binary compatibility--assumes one level of nesting.
*/
if (is64bit) {
if ((error = copyin(fsmountargs, (caddr_t)&devpath, sizeof(devpath)))) {
return error;
}
} else {
user32_addr_t tmp;
if ((error = copyin(fsmountargs, (caddr_t)&tmp, sizeof(tmp)))) {
return error;
}
/* munge into LP64 addr */
devpath = CAST_USER_ADDR_T(tmp);
}
height = 0;
flags = 0;
}
if (flags != 0) {
IMGSRC_DEBUG("%s: Got nonzero flags.\n", __FUNCTION__);
return EINVAL;
}
error = get_imgsrc_rootvnode(height, &rvp);
if (error != 0) {
IMGSRC_DEBUG("getting old root vnode failed with %d\n", error);
return error;
}
IMGSRC_DEBUG("got old root vnode\n");
old_mntonname = zalloc_flags(ZV_NAMEI, Z_WAITOK);
/* Can only move once */
mp = vnode_mount(rvp);
if ((mp->mnt_kern_flag & MNTK_HAS_MOVED) == MNTK_HAS_MOVED) {
IMGSRC_DEBUG("Already moved.\n");
error = EBUSY;
goto out0;
}
IMGSRC_DEBUG("moving rvp: fsname = %s\n", mp->mnt_vtable->vfc_name);
IMGSRC_DEBUG("Starting updated.\n");
/* Get exclusive rwlock on mount, authorize update on mp */
error = mount_begin_update(mp, ctx, 0);
if (error != 0) {
IMGSRC_DEBUG("Starting updated failed with %d\n", error);
goto out0;
}
/*
* It can only be moved once. Flag is set under the rwlock,
* so we're now safe to proceed.
*/
if ((mp->mnt_kern_flag & MNTK_HAS_MOVED) == MNTK_HAS_MOVED) {
IMGSRC_DEBUG("Already moved [2]\n");
goto out1;
}
IMGSRC_DEBUG("Preparing coveredvp.\n");
/* Mark covered vnode as mount in progress, authorize placing mount on top */
error = prepare_coveredvp(vp, ctx, cnp, fsname, 0);
if (error != 0) {
IMGSRC_DEBUG("Preparing coveredvp failed with %d.\n", error);
goto out1;
}
IMGSRC_DEBUG("Covered vp OK.\n");
/* Sanity check the name caller has provided */
vfsp = mp->mnt_vtable;
if (strncmp(vfsp->vfc_name, fsname, MFSNAMELEN) != 0) {
IMGSRC_DEBUG("Wrong fs name: actual = %s, expected = %s\n",
vfsp->vfc_name, fsname);
error = EINVAL;
goto out2;
}
/* Check the device vnode and update mount-from name, for local filesystems */
if (vfsp->vfc_vfsflags & VFC_VFSLOCALARGS) {
IMGSRC_DEBUG("Local, doing device validation.\n");
if (devpath != USER_ADDR_NULL) {
error = authorize_devpath_and_update_mntfromname(mp, devpath, &devvp, ctx);
if (error) {
IMGSRC_DEBUG("authorize_devpath_and_update_mntfromname() failed.\n");
goto out2;
}
vnode_put(devvp);
}
}
/*
* Place mp on top of vnode, ref the vnode, call checkdirs(),
* and increment the name cache's mount generation
*/
IMGSRC_DEBUG("About to call place_mount_and_checkdirs().\n");
error = place_mount_and_checkdirs(mp, vp, ctx);
if (error != 0) {
goto out2;
}
placed = TRUE;
strlcpy(old_mntonname, mp->mnt_vfsstat.f_mntonname, MAXPATHLEN);
strlcpy(mp->mnt_vfsstat.f_mntonname, cnp->cn_pnbuf, MAXPATHLEN);
/* Forbid future moves */
mount_lock(mp);
mp->mnt_kern_flag |= MNTK_HAS_MOVED;
mount_unlock(mp);
/* Finally, add to mount list, completely ready to go */
if (mount_list_add(mp) != 0) {
/*
* The system is shutting down trying to umount
* everything, so fail with a plausible errno.
*/
error = EBUSY;
goto out3;
}
mount_end_update(mp);
vnode_put(rvp);
zfree(ZV_NAMEI, old_mntonname);
vfs_notify_mount(pvp);
return 0;
out3:
strlcpy(mp->mnt_vfsstat.f_mntonname, old_mntonname, MAXPATHLEN);
mount_lock(mp);
mp->mnt_kern_flag &= ~(MNTK_HAS_MOVED);
mount_unlock(mp);
out2:
/*
* Placing the mp on the vnode clears VMOUNT,
* so cleanup is different after that point
*/
if (placed) {
/* Rele the vp, clear VMOUNT and v_mountedhere */
undo_place_on_covered_vp(mp, vp);
} else {
vnode_lock_spin(vp);
CLR(vp->v_flag, VMOUNT);
vnode_unlock(vp);
}
out1:
mount_end_update(mp);
out0:
vnode_put(rvp);
zfree(ZV_NAMEI, old_mntonname);
return error;
}
#endif /* CONFIG_IMGSRC_ACCESS */
void
enablequotas(struct mount *mp, vfs_context_t ctx)
{
struct nameidata qnd;
int type;
char qfpath[MAXPATHLEN];
const char *qfname = QUOTAFILENAME;
const char *qfopsname = QUOTAOPSNAME;
const char *qfextension[] = INITQFNAMES;
/* XXX Shoulkd be an MNTK_ flag, instead of strncmp()'s */
if (strncmp(mp->mnt_vfsstat.f_fstypename, "hfs", sizeof("hfs")) != 0) {
return;
}
/*
* Enable filesystem disk quotas if necessary.
* We ignore errors as this should not interfere with final mount
*/
for (type = 0; type < MAXQUOTAS; type++) {
snprintf(qfpath, sizeof(qfpath), "%s/%s.%s", mp->mnt_vfsstat.f_mntonname, qfopsname, qfextension[type]);
NDINIT(&qnd, LOOKUP, OP_MOUNT, FOLLOW, UIO_SYSSPACE,
CAST_USER_ADDR_T(qfpath), ctx);
if (namei(&qnd) != 0) {
continue; /* option file to trigger quotas is not present */
}
vnode_put(qnd.ni_vp);
nameidone(&qnd);
snprintf(qfpath, sizeof(qfpath), "%s/%s.%s", mp->mnt_vfsstat.f_mntonname, qfname, qfextension[type]);
(void) VFS_QUOTACTL(mp, QCMD(Q_QUOTAON, type), 0, qfpath, ctx);
}
return;
}
static int
checkdirs_callback(proc_t p, void * arg)
{
struct cdirargs *cdrp = (struct cdirargs *)arg;
vnode_t olddp = cdrp->olddp;
vnode_t newdp = cdrp->newdp;
struct filedesc *fdp = &p->p_fd;
vnode_t new_cvp = newdp;
vnode_t new_rvp = newdp;
vnode_t old_cvp = NULL;
vnode_t old_rvp = NULL;
/*
* XXX Also needs to iterate each thread in the process to see if it
* XXX is using a per-thread current working directory, and, if so,
* XXX update that as well.
*/
/*
* First, with the proc_fdlock held, check to see if we will need
* to do any work. If not, we will get out fast.
*/
proc_fdlock(p);
if (fdp->fd_cdir != olddp && fdp->fd_rdir != olddp) {
proc_fdunlock(p);
return PROC_RETURNED;
}
proc_fdunlock(p);
/*
* Ok, we will have to do some work. Always take two refs
* because we might need that many. We'll dispose of whatever
* we ended up not using.
*/
if (vnode_ref(newdp) != 0) {
return PROC_RETURNED;
}
if (vnode_ref(newdp) != 0) {
vnode_rele(newdp);
return PROC_RETURNED;
}
proc_dirs_lock_exclusive(p);
/*
* Now do the work. Note: we dropped the proc_fdlock, so we
* have to do all of the checks again.
*/
proc_fdlock(p);
if (fdp->fd_cdir == olddp) {
old_cvp = olddp;
fdp->fd_cdir = newdp;
new_cvp = NULL;
}
if (fdp->fd_rdir == olddp) {
old_rvp = olddp;
fdp->fd_rdir = newdp;
new_rvp = NULL;
}
proc_fdunlock(p);
proc_dirs_unlock_exclusive(p);
/*
* Dispose of any references that are no longer needed.
*/
if (old_cvp != NULL) {
vnode_rele(old_cvp);
}
if (old_rvp != NULL) {
vnode_rele(old_rvp);
}
if (new_cvp != NULL) {
vnode_rele(new_cvp);
}
if (new_rvp != NULL) {
vnode_rele(new_rvp);
}
return PROC_RETURNED;
}
/*
* Scan all active processes to see if any of them have a current
* or root directory onto which the new filesystem has just been
* mounted. If so, replace them with the new mount point.
*/
static int
checkdirs(vnode_t olddp, vfs_context_t ctx)
{
vnode_t newdp;
vnode_t tvp;
int err;
struct cdirargs cdr;
if (olddp->v_usecount == 1) {
return 0;
}
err = VFS_ROOT(olddp->v_mountedhere, &newdp, ctx);
if (err != 0) {
#if DIAGNOSTIC
panic("mount: lost mount: error %d", err);
#endif
return err;
}
cdr.olddp = olddp;
cdr.newdp = newdp;
/* do not block for exec/fork trans as the vp in cwd & rootdir are not changing */
proc_iterate(PROC_ALLPROCLIST | PROC_NOWAITTRANS, checkdirs_callback, (void *)&cdr, NULL, NULL);
if (rootvnode == olddp) {
vnode_ref(newdp);
lck_rw_lock_exclusive(&rootvnode_rw_lock);
tvp = rootvnode;
rootvnode = newdp;
lck_rw_unlock_exclusive(&rootvnode_rw_lock);
vnode_rele(tvp);
}
vnode_put(newdp);
return 0;
}
#define ROLE_ACCOUNT_UNMOUNT_ENTITLEMENT \
"com.apple.private.vfs.role-account-unmount"
/*
* Unmount a file system.
*
* Note: unmount takes a path to the vnode mounted on as argument,
* not special file (as before).
*/
/* ARGSUSED */
int
unmount(__unused proc_t p, struct unmount_args *uap, __unused int32_t *retval)
{
vnode_t vp;
struct mount *mp;
int error;
struct nameidata nd;
vfs_context_t ctx;
/*
* If the process has the entitlement, use the kernel's context when
* performing lookup on the mount path as the process might lack proper
* permission to access the directory.
*/
ctx = IOCurrentTaskHasEntitlement(ROLE_ACCOUNT_UNMOUNT_ENTITLEMENT) ?
vfs_context_kernel() : vfs_context_current();
NDINIT(&nd, LOOKUP, OP_UNMOUNT, FOLLOW | AUDITVNPATH1,
UIO_USERSPACE, uap->path, ctx);
error = namei(&nd);
if (error) {
return error;
}
vp = nd.ni_vp;
mp = vp->v_mount;
nameidone(&nd);
/*
* Must be the root of the filesystem
*/
if ((vp->v_flag & VROOT) == 0) {
vnode_put(vp);
return EINVAL;
}
#if CONFIG_MACF
error = mac_mount_check_umount(ctx, mp);
if (error != 0) {
vnode_put(vp);
return error;
}
#endif
mount_ref(mp, 0);
vnode_put(vp);
/* safedounmount consumes the mount ref */
return safedounmount(mp, uap->flags, ctx);
}
int
vfs_unmountbyfsid(fsid_t *fsid, int flags, vfs_context_t ctx)
{
mount_t mp;
mp = mount_list_lookupby_fsid(fsid, 0, 1);
if (mp == (mount_t)0) {
return ENOENT;
}
mount_ref(mp, 0);
mount_iterdrop(mp);
/* safedounmount consumes the mount ref */
return safedounmount(mp, flags, ctx);
}
/*
* The mount struct comes with a mount ref which will be consumed.
* Do the actual file system unmount, prevent some common foot shooting.
*/
int
safedounmount(struct mount *mp, int flags, vfs_context_t ctx)
{
int error;
proc_t p = vfs_context_proc(ctx);
/*
* If the file system is not responding and MNT_NOBLOCK
* is set and not a forced unmount then return EBUSY.
*/
if ((mp->mnt_kern_flag & MNT_LNOTRESP) &&
(flags & MNT_NOBLOCK) && ((flags & MNT_FORCE) == 0)) {
error = EBUSY;
goto out;
}
/*
* Skip authorization in two cases:
* - If the process running the unmount has ROLE_ACCOUNT_UNMOUNT_ENTITLEMENT.
* This entitlement allows non-root processes unmount volumes mounted by
* other processes.
* - If the mount is tagged as permissive and this is not a forced-unmount
* attempt.
*/
if (!IOCurrentTaskHasEntitlement(ROLE_ACCOUNT_UNMOUNT_ENTITLEMENT) &&
(!(((mp->mnt_kern_flag & MNTK_PERMIT_UNMOUNT) != 0) && ((flags & MNT_FORCE) == 0)))) {
/*
* Only root, or the user that did the original mount is
* permitted to unmount this filesystem.
*/
if ((mp->mnt_vfsstat.f_owner != kauth_cred_getuid(kauth_cred_get())) &&
(error = suser(kauth_cred_get(), &p->p_acflag))) {
goto out;
}
}
/*
* Don't allow unmounting the root file system, or other volumes
* associated with it (for example, the associated VM or DATA mounts) .
*/
if ((mp->mnt_flag & MNT_ROOTFS) || (mp->mnt_kern_flag & MNTK_SYSTEM)) {
if (!(mp->mnt_flag & MNT_ROOTFS)) {
printf("attempt to unmount a system mount (%s), will return EBUSY\n",
mp->mnt_vfsstat.f_mntonname);
}
error = EBUSY; /* the root (or associated volumes) is always busy */
goto out;
}
/*
* If the mount is providing the root filesystem's disk image
* (i.e. imageboot), don't allow unmounting
*/
if (mp->mnt_kern_flag & MNTK_BACKS_ROOT) {
error = EBUSY;
goto out;
}
return dounmount(mp, flags, 1, ctx);
out:
mount_drop(mp, 0);
return error;
}
/*
* Do the actual file system unmount.
*/
int
dounmount(struct mount *mp, int flags, int withref, vfs_context_t ctx)
{
vnode_t coveredvp = (vnode_t)0;
int error;
int needwakeup = 0;
int forcedunmount = 0;
int lflags = 0;
struct vnode *devvp = NULLVP;
#if CONFIG_TRIGGERS
proc_t p = vfs_context_proc(ctx);
int did_vflush = 0;
int pflags_save = 0;
#endif /* CONFIG_TRIGGERS */
#if CONFIG_FSE
if (!(flags & MNT_FORCE)) {
fsevent_unmount(mp, ctx); /* has to come first! */
}
#endif
mount_lock(mp);
/*
* If already an unmount in progress just return EBUSY.
* Even a forced unmount cannot override.
*/
if (mp->mnt_lflag & (MNT_LUNMOUNT | MNT_LMOUNT)) {
if (withref != 0) {
mount_drop(mp, 1);
}
mount_unlock(mp);
return EBUSY;
}
if (flags & MNT_FORCE) {
forcedunmount = 1;
mp->mnt_lflag |= MNT_LFORCE;
}
#if CONFIG_TRIGGERS
if (flags & MNT_NOBLOCK && p != kernproc) {
pflags_save = OSBitOrAtomic(P_NOREMOTEHANG, &p->p_flag);
}
#endif
mp->mnt_kern_flag |= MNTK_UNMOUNT;
mp->mnt_lflag |= MNT_LUNMOUNT;
mp->mnt_flag &= ~MNT_ASYNC;
/*
* anyone currently in the fast path that
* trips over the cached rootvp will be
* dumped out and forced into the slow path
* to regenerate a new cached value
*/
mp->mnt_realrootvp = NULLVP;
mount_unlock(mp);
if (forcedunmount && (flags & MNT_LNOSUB) == 0) {
/*
* Force unmount any mounts in this filesystem.
* If any unmounts fail - just leave them dangling.
* Avoids recursion.
*/
(void) dounmount_submounts(mp, flags | MNT_LNOSUB, ctx);
}
/*
* taking the name_cache_lock exclusively will
* insure that everyone is out of the fast path who
* might be trying to use a now stale copy of
* vp->v_mountedhere->mnt_realrootvp
* bumping mount_generation causes the cached values
* to be invalidated
*/
name_cache_lock();
mount_generation++;
name_cache_unlock();
lck_rw_lock_exclusive(&mp->mnt_rwlock);
if (withref != 0) {
mount_drop(mp, 0);
}
error = 0;
if (forcedunmount == 0) {
ubc_umount(mp); /* release cached vnodes */
if ((mp->mnt_flag & MNT_RDONLY) == 0) {
error = VFS_SYNC(mp, MNT_WAIT, ctx);
if (error) {
mount_lock(mp);
mp->mnt_kern_flag &= ~MNTK_UNMOUNT;
mp->mnt_lflag &= ~MNT_LUNMOUNT;
mp->mnt_lflag &= ~MNT_LFORCE;
goto out;
}
}
}
IOBSDMountChange(mp, kIOMountChangeUnmount);
#if CONFIG_TRIGGERS
vfs_nested_trigger_unmounts(mp, flags, ctx);
did_vflush = 1;
#endif
if (forcedunmount) {
lflags |= FORCECLOSE;
}
error = vflush(mp, NULLVP, SKIPSWAP | SKIPSYSTEM | SKIPROOT | lflags);
if ((forcedunmount == 0) && error) {
mount_lock(mp);
mp->mnt_kern_flag &= ~MNTK_UNMOUNT;
mp->mnt_lflag &= ~MNT_LUNMOUNT;
mp->mnt_lflag &= ~MNT_LFORCE;
goto out;
}
/* make sure there are no one in the mount iterations or lookup */
mount_iterdrain(mp);
error = VFS_UNMOUNT(mp, flags, ctx);
if (error) {
mount_iterreset(mp);
mount_lock(mp);
mp->mnt_kern_flag &= ~MNTK_UNMOUNT;
mp->mnt_lflag &= ~MNT_LUNMOUNT;
mp->mnt_lflag &= ~MNT_LFORCE;
goto out;
}
/* increment the operations count */
if (!error) {
OSAddAtomic(1, &vfs_nummntops);
}
if (mp->mnt_devvp && mp->mnt_vtable->vfc_vfsflags & VFC_VFSLOCALARGS) {
/* hold an io reference and drop the usecount before close */
devvp = mp->mnt_devvp;
vnode_getalways(devvp);
vnode_rele(devvp);
VNOP_CLOSE(devvp, mp->mnt_flag & MNT_RDONLY ? FREAD : FREAD | FWRITE,
ctx);
vnode_clearmountedon(devvp);
vnode_put(devvp);
}
lck_rw_done(&mp->mnt_rwlock);
mount_list_remove(mp);
lck_rw_lock_exclusive(&mp->mnt_rwlock);
/* mark the mount point hook in the vp but not drop the ref yet */
if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) {
/*
* The covered vnode needs special handling. Trying to get an
* iocount must not block here as this may lead to deadlocks
* if the Filesystem to which the covered vnode belongs is
* undergoing forced unmounts. Since we hold a usecount, the
* vnode cannot be reused (it can, however, still be terminated)
*/
vnode_getalways(coveredvp);
vnode_lock_spin(coveredvp);
mp->mnt_crossref++;
coveredvp->v_mountedhere = (struct mount *)0;
CLR(coveredvp->v_flag, VMOUNT | VMOUNTEDHERE);
vnode_unlock(coveredvp);
vnode_put(coveredvp);
}
mount_list_lock();
mp->mnt_vtable->vfc_refcount--;
mount_list_unlock();
cache_purgevfs(mp); /* remove cache entries for this file sys */
vfs_event_signal(NULL, VQ_UNMOUNT, (intptr_t)NULL);
mount_lock(mp);
mp->mnt_lflag |= MNT_LDEAD;
if (mp->mnt_lflag & MNT_LWAIT) {
/*
* do the wakeup here
* in case we block in mount_refdrain
* which will drop the mount lock
* and allow anyone blocked in vfs_busy
* to wakeup and see the LDEAD state
*/
mp->mnt_lflag &= ~MNT_LWAIT;
wakeup((caddr_t)mp);
}
mount_refdrain(mp);
/* free disk_conditioner_info structure for this mount */
disk_conditioner_unmount(mp);
out:
if (mp->mnt_lflag & MNT_LWAIT) {
mp->mnt_lflag &= ~MNT_LWAIT;
needwakeup = 1;
}
#if CONFIG_TRIGGERS
if (flags & MNT_NOBLOCK && p != kernproc) {
// Restore P_NOREMOTEHANG bit to its previous value
if ((pflags_save & P_NOREMOTEHANG) == 0) {
OSBitAndAtomic(~((uint32_t) P_NOREMOTEHANG), &p->p_flag);
}
}
/*
* Callback and context are set together under the mount lock, and
* never cleared, so we're safe to examine them here, drop the lock,
* and call out.
*/
if (mp->mnt_triggercallback != NULL) {
mount_unlock(mp);
if (error == 0) {
mp->mnt_triggercallback(mp, VTC_RELEASE, mp->mnt_triggerdata, ctx);
} else if (did_vflush) {
mp->mnt_triggercallback(mp, VTC_REPLACE, mp->mnt_triggerdata, ctx);
}
} else {
mount_unlock(mp);
}
#else
mount_unlock(mp);
#endif /* CONFIG_TRIGGERS */
lck_rw_done(&mp->mnt_rwlock);
if (needwakeup) {
wakeup((caddr_t)mp);
}
if (!error) {
if ((coveredvp != NULLVP)) {
vnode_t pvp = NULLVP;
/*
* The covered vnode needs special handling. Trying to
* get an iocount must not block here as this may lead
* to deadlocks if the Filesystem to which the covered
* vnode belongs is undergoing forced unmounts. Since we
* hold a usecount, the vnode cannot be reused
* (it can, however, still be terminated).
*/
vnode_getalways(coveredvp);
mount_dropcrossref(mp, coveredvp, 0);
/*
* We'll _try_ to detect if this really needs to be
* done. The coveredvp can only be in termination (or
* terminated) if the coveredvp's mount point is in a
* forced unmount (or has been) since we still hold the
* ref.
*/
if (!vnode_isrecycled(coveredvp)) {
pvp = vnode_getparent(coveredvp);
#if CONFIG_TRIGGERS
if (coveredvp->v_resolve) {
vnode_trigger_rearm(coveredvp, ctx);
}
#endif
}
vnode_rele(coveredvp);
vnode_put(coveredvp);
coveredvp = NULLVP;
if (pvp) {
lock_vnode_and_post(pvp, NOTE_WRITE);
vnode_put(pvp);
}
} else if (mp->mnt_flag & MNT_ROOTFS) {
if (nc_smr_enabled) {
vfs_smr_synchronize();
}
mount_lock_destroy(mp);
#if CONFIG_MACF
mac_mount_label_destroy(mp);
#endif
zfree(mount_zone, mp);
} else {
panic("dounmount: no coveredvp");
}
}
return error;
}
/*
* Unmount any mounts in this filesystem.
*/
void
dounmount_submounts(struct mount *mp, int flags, vfs_context_t ctx)
{
mount_t smp;
fsid_t *fsids, fsid;
int fsids_sz;
int count = 0, i, m = 0;
vnode_t vp;
mount_list_lock();
// Get an array to hold the submounts fsids.
TAILQ_FOREACH(smp, &mountlist, mnt_list)
count++;
fsids_sz = count * sizeof(fsid_t);
fsids = kalloc_data(fsids_sz, Z_NOWAIT);
if (fsids == NULL) {
mount_list_unlock();
goto out;
}
fsids[0] = mp->mnt_vfsstat.f_fsid; // Prime the pump
/*
* Fill the array with submount fsids.
* Since mounts are always added to the tail of the mount list, the
* list is always in mount order.
* For each mount check if the mounted-on vnode belongs to a
* mount that's already added to our array of mounts to be unmounted.
*/
for (smp = TAILQ_NEXT(mp, mnt_list); smp; smp = TAILQ_NEXT(smp, mnt_list)) {
vp = smp->mnt_vnodecovered;
if (vp == NULL) {
continue;
}
fsid = vnode_mount(vp)->mnt_vfsstat.f_fsid; // Underlying fsid
for (i = 0; i <= m; i++) {
if (fsids[i].val[0] == fsid.val[0] &&
fsids[i].val[1] == fsid.val[1]) {
fsids[++m] = smp->mnt_vfsstat.f_fsid;
break;
}
}
}
mount_list_unlock();
// Unmount the submounts in reverse order. Ignore errors.
for (i = m; i > 0; i--) {
smp = mount_list_lookupby_fsid(&fsids[i], 0, 1);
if (smp) {
mount_ref(smp, 0);
mount_iterdrop(smp);
(void) dounmount(smp, flags, 1, ctx);
}
}
out:
kfree_data(fsids, fsids_sz);
}
void
mount_dropcrossref(mount_t mp, vnode_t dp, int need_put)
{
vnode_hold(dp);
vnode_lock(dp);
mp->mnt_crossref--;
if (mp->mnt_crossref < 0) {
panic("mount cross refs -ve");
}
if ((mp != dp->v_mountedhere) && (mp->mnt_crossref == 0)) {
if (need_put) {
vnode_put_locked(dp);
}
vnode_drop_and_unlock(dp);
if (nc_smr_enabled) {
vfs_smr_synchronize();
}
mount_lock_destroy(mp);
#if CONFIG_MACF
mac_mount_label_destroy(mp);
#endif
zfree(mount_zone, mp);
return;
}
if (need_put) {
vnode_put_locked(dp);
}
vnode_drop_and_unlock(dp);
}
/*
* Sync each mounted filesystem.
*/
#if DIAGNOSTIC
int syncprt = 0;
#endif
int print_vmpage_stat = 0;
/*
* sync_callback: simple wrapper that calls VFS_SYNC() on volumes
* mounted read-write with the passed waitfor value.
*
* Parameters: mp mount-point descriptor per mounted file-system instance.
* arg user argument (please see below)
*
* User argument is a pointer to 32 bit unsigned integer which describes the
* type of waitfor value to set for calling VFS_SYNC(). If user argument is
* passed as NULL, VFS_SYNC() is called with MNT_NOWAIT set as the default
* waitfor value.
*
* Returns: VFS_RETURNED
*/
static int
sync_callback(mount_t mp, void *arg)
{
if ((mp->mnt_flag & MNT_RDONLY) == 0) {
int asyncflag = mp->mnt_flag & MNT_ASYNC;
unsigned waitfor = MNT_NOWAIT;
if (arg) {
waitfor = *(uint32_t*)arg;
}
/* Sanity check for flags - these are the only valid combinations for the flag bits*/
if (waitfor != MNT_WAIT &&
waitfor != (MNT_WAIT | MNT_VOLUME) &&
waitfor != MNT_NOWAIT &&
waitfor != (MNT_NOWAIT | MNT_VOLUME) &&
waitfor != MNT_DWAIT &&
waitfor != (MNT_DWAIT | MNT_VOLUME)) {
panic("Passed inappropriate waitfor %u to "
"sync_callback()", waitfor);
}
mp->mnt_flag &= ~MNT_ASYNC;
(void)VFS_SYNC(mp, waitfor, vfs_context_kernel());
if (asyncflag) {
mp->mnt_flag |= MNT_ASYNC;
}
}
return VFS_RETURNED;
}
/* ARGSUSED */
int
sync(__unused proc_t p, __unused struct sync_args *uap, __unused int32_t *retval)
{
vfs_iterate(LK_NOWAIT, sync_callback, NULL);
if (print_vmpage_stat) {
vm_countdirtypages();
}
#if DIAGNOSTIC
if (syncprt) {
vfs_bufstats();
}
#endif /* DIAGNOSTIC */
return 0;
}
typedef enum {
SYNC_ALL = 0,
SYNC_ONLY_RELIABLE_MEDIA = 1,
SYNC_ONLY_UNRELIABLE_MEDIA = 2
} sync_type_t;
static int
sync_internal_callback(mount_t mp, void *arg)
{
if (arg) {
int is_reliable = !(mp->mnt_kern_flag & MNTK_VIRTUALDEV) &&
(mp->mnt_flag & MNT_LOCAL);
sync_type_t sync_type = *((sync_type_t *)arg);
if ((sync_type == SYNC_ONLY_RELIABLE_MEDIA) && !is_reliable) {
return VFS_RETURNED;
} else if ((sync_type == SYNC_ONLY_UNRELIABLE_MEDIA) && is_reliable) {
return VFS_RETURNED;
}
}
(void)sync_callback(mp, NULL);
return VFS_RETURNED;
}
int sync_thread_state = 0;
int sync_timeout_seconds = 5;
#define SYNC_THREAD_RUN 0x0001
#define SYNC_THREAD_RUNNING 0x0002
#if CONFIG_PHYS_WRITE_ACCT
thread_t pm_sync_thread;
#endif /* CONFIG_PHYS_WRITE_ACCT */
static void
sync_thread(__unused void *arg, __unused wait_result_t wr)
{
sync_type_t sync_type;
#if CONFIG_PHYS_WRITE_ACCT
pm_sync_thread = current_thread();
#endif /* CONFIG_PHYS_WRITE_ACCT */
lck_mtx_lock(&sync_mtx_lck);
while (sync_thread_state & SYNC_THREAD_RUN) {
sync_thread_state &= ~SYNC_THREAD_RUN;
lck_mtx_unlock(&sync_mtx_lck);
sync_type = SYNC_ONLY_RELIABLE_MEDIA;
vfs_iterate(LK_NOWAIT, sync_internal_callback, &sync_type);
sync_type = SYNC_ONLY_UNRELIABLE_MEDIA;
vfs_iterate(LK_NOWAIT, sync_internal_callback, &sync_type);
lck_mtx_lock(&sync_mtx_lck);
}
/*
* This wakeup _has_ to be issued before the lock is released otherwise
* we may end up waking up a thread in sync_internal which is
* expecting a wakeup from a thread it just created and not from this
* thread which is about to exit.
*/
wakeup(&sync_thread_state);
sync_thread_state &= ~SYNC_THREAD_RUNNING;
#if CONFIG_PHYS_WRITE_ACCT
pm_sync_thread = NULL;
#endif /* CONFIG_PHYS_WRITE_ACCT */
lck_mtx_unlock(&sync_mtx_lck);
if (print_vmpage_stat) {
vm_countdirtypages();
}
#if DIAGNOSTIC
if (syncprt) {
vfs_bufstats();
}
#endif /* DIAGNOSTIC */
}
struct timeval sync_timeout_last_print = {.tv_sec = 0, .tv_usec = 0};
/*
* An in-kernel sync for power management to call.
* This function always returns within sync_timeout seconds.
*/
__private_extern__ int
sync_internal(void)
{
thread_t thd = NULL;
int error;
int thread_created = FALSE;
struct timespec ts = {.tv_sec = sync_timeout_seconds, .tv_nsec = 0};
lck_mtx_lock(&sync_mtx_lck);
sync_thread_state |= SYNC_THREAD_RUN;
if (!(sync_thread_state & SYNC_THREAD_RUNNING)) {
int kr;
sync_thread_state |= SYNC_THREAD_RUNNING;
kr = kernel_thread_start(sync_thread, NULL, &thd);
if (kr != KERN_SUCCESS) {
sync_thread_state &= ~SYNC_THREAD_RUNNING;
lck_mtx_unlock(&sync_mtx_lck);
printf("sync_thread failed\n");
return 0;
}
thread_created = TRUE;
}
error = msleep((caddr_t)&sync_thread_state, &sync_mtx_lck,
(PVFS | PDROP | PCATCH), "sync_thread", &ts);
if (error) {
struct timeval now;
microtime(&now);
if (now.tv_sec - sync_timeout_last_print.tv_sec > 120) {
printf("sync timed out: %d sec\n", sync_timeout_seconds);
sync_timeout_last_print.tv_sec = now.tv_sec;
}
}
if (thread_created) {
thread_deallocate(thd);
}
return 0;
} /* end of sync_internal call */
/*
* Change filesystem quotas.
*/
#if QUOTA
int
quotactl(proc_t p, struct quotactl_args *uap, __unused int32_t *retval)
{
struct mount *mp;
int error, quota_cmd, quota_status = 0;
caddr_t datap;
size_t fnamelen;
struct nameidata nd;
vfs_context_t ctx = vfs_context_current();
struct dqblk my_dqblk = {};
AUDIT_ARG(uid, uap->uid);
AUDIT_ARG(cmd, uap->cmd);
NDINIT(&nd, LOOKUP, OP_LOOKUP, FOLLOW | AUDITVNPATH1, UIO_USERSPACE,
uap->path, ctx);
error = namei(&nd);
if (error) {
return error;
}
mp = nd.ni_vp->v_mount;
mount_ref(mp, 0);
vnode_put(nd.ni_vp);
nameidone(&nd);
#if CONFIG_MACF
error = mac_mount_check_quotactl(ctx, mp, uap->cmd, uap->uid);
if (error != 0) {
goto out;
}
#endif
/* copyin any data we will need for downstream code */
quota_cmd = uap->cmd >> SUBCMDSHIFT;
switch (quota_cmd) {
case Q_QUOTAON:
/* uap->arg specifies a file from which to take the quotas */
fnamelen = MAXPATHLEN;
datap = zalloc(ZV_NAMEI);
error = copyinstr(uap->arg, datap, MAXPATHLEN, &fnamelen);
break;
case Q_GETQUOTA:
/* uap->arg is a pointer to a dqblk structure. */
datap = (caddr_t) &my_dqblk;
break;
case Q_SETQUOTA:
case Q_SETUSE:
/* uap->arg is a pointer to a dqblk structure. */
datap = (caddr_t) &my_dqblk;
if (proc_is64bit(p)) {
struct user_dqblk my_dqblk64;
error = copyin(uap->arg, (caddr_t)&my_dqblk64, sizeof(my_dqblk64));
if (error == 0) {
munge_dqblk(&my_dqblk, &my_dqblk64, FALSE);
}
} else {
error = copyin(uap->arg, (caddr_t)&my_dqblk, sizeof(my_dqblk));
}
break;
case Q_QUOTASTAT:
/* uap->arg is a pointer to an integer */
datap = (caddr_t) &quota_status;
break;
default:
datap = NULL;
break;
} /* switch */
if (error == 0) {
error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, datap, ctx);
}
switch (quota_cmd) {
case Q_QUOTAON:
if (datap != NULL) {
zfree(ZV_NAMEI, datap);
}
break;
case Q_GETQUOTA:
/* uap->arg is a pointer to a dqblk structure we need to copy out to */
if (error == 0) {
if (proc_is64bit(p)) {
struct user_dqblk my_dqblk64;
memset(&my_dqblk64, 0, sizeof(my_dqblk64));
munge_dqblk(&my_dqblk, &my_dqblk64, TRUE);
error = copyout((caddr_t)&my_dqblk64, uap->arg, sizeof(my_dqblk64));
} else {
error = copyout(datap, uap->arg, sizeof(struct dqblk));
}
}
break;
case Q_QUOTASTAT:
/* uap->arg is a pointer to an integer */
if (error == 0) {
error = copyout(datap, uap->arg, sizeof(quota_status));
}
break;
default:
break;
} /* switch */
out:
mount_drop(mp, 0);
return error;
}
#else
int
quotactl(__unused proc_t p, __unused struct quotactl_args *uap, __unused int32_t *retval)
{
return EOPNOTSUPP;
}
#endif /* QUOTA */
static int
statfs_internal(proc_t p, struct mount *mp, user_addr_t bufp)
{
int error;
vfs_context_t ctx = vfs_context_current();
#if CONFIG_MACF
error = mac_mount_check_stat(ctx, mp);
if (error != 0) {
return error;
}
#endif
error = vfs_update_vfsstat(mp, ctx, VFS_USER_EVENT);
if (error != 0) {
return error;
}
return munge_statfs(mp, &mp->mnt_vfsstat, bufp, NULL, IS_64BIT_PROCESS(p), TRUE);
}
/*
* Get filesystem statistics.
*
* Returns: 0 Success
* namei:???
* vfs_update_vfsstat:???
* munge_statfs:EFAULT
*/
/* ARGSUSED */
int
statfs(proc_t p, struct statfs_args *uap, __unused int32_t *retval)
{
int error;
struct mount *mp;
struct nameidata nd;
vfs_context_t ctx = vfs_context_current();
vnode_t vp;
NDINIT(&nd, LOOKUP, OP_STATFS, FOLLOW | AUDITVNPATH1,
UIO_USERSPACE, uap->path, ctx);
error = namei(&nd);
if (error != 0) {
return error;
}
vp = nd.ni_vp;
mp = vp->v_mount;
nameidone(&nd);
error = statfs_internal(p, mp, uap->buf);
vnode_put(vp);
return error;
}
/*
* Get filesystem statistics.
*/
/* ARGSUSED */
int
fstatfs(proc_t p, struct fstatfs_args *uap, __unused int32_t *retval)
{
int error;
vnode_t vp = NULL;
struct mount *mp;
AUDIT_ARG(fd, uap->fd);
if ((error = file_vnode(uap->fd, &vp)) ||
(error = vnode_getwithref(vp))) {
goto out;
}
AUDIT_ARG(vnpath_withref, vp, ARG_VNODE1);
mp = vp->v_mount;
if (!mp) {
error = EBADF;
goto out_vnode;
}
error = statfs_internal(p, mp, uap->buf);
out_vnode:
vnode_put(vp);
out:
if (vp != NULL) {
file_drop(uap->fd);
}
return error;
}
void
vfs_get_statfs64(struct mount *mp, struct statfs64 *sfs)
{
struct vfsstatfs *vsfs = &mp->mnt_vfsstat;
bzero(sfs, sizeof(*sfs));
sfs->f_bsize = vsfs->f_bsize;
sfs->f_iosize = (int32_t)vsfs->f_iosize;
sfs->f_blocks = vsfs->f_blocks;
sfs->f_bfree = vsfs->f_bfree;
sfs->f_bavail = vsfs->f_bavail;
sfs->f_files = vsfs->f_files;
sfs->f_ffree = vsfs->f_ffree;
sfs->f_fsid = vsfs->f_fsid;
sfs->f_owner = vsfs->f_owner;
sfs->f_type = mp->mnt_vtable->vfc_typenum;
sfs->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
sfs->f_fssubtype = vsfs->f_fssubtype;
sfs->f_flags_ext = 0;
if (mp->mnt_kern_flag & MNTK_SYSTEMDATA) {
sfs->f_flags_ext |= MNT_EXT_ROOT_DATA_VOL;
}
if (mp->mnt_kern_flag & MNTK_FSKIT) {
sfs->f_flags_ext |= MNT_EXT_FSKIT;
}
vfs_getfstypename(mp, sfs->f_fstypename, MFSTYPENAMELEN);
strlcpy(&sfs->f_mntonname[0], &vsfs->f_mntonname[0], MAXPATHLEN);
strlcpy(&sfs->f_mntfromname[0], &vsfs->f_mntfromname[0], MAXPATHLEN);
}
/*
* Get file system statistics in 64-bit mode
*/
int
statfs64(__unused struct proc *p, struct statfs64_args *uap, __unused int32_t *retval)
{
struct mount *mp;
int error;
struct nameidata *ndp;
struct statfs64 *sfsp;
vfs_context_t ctxp = vfs_context_current();
vnode_t vp;
struct {
struct nameidata nd;
struct statfs64 sfs;
} *__nameidata_statfs64;
__nameidata_statfs64 = kalloc_type(typeof(*__nameidata_statfs64),
Z_WAITOK);
ndp = &__nameidata_statfs64->nd;
NDINIT(ndp, LOOKUP, OP_STATFS, FOLLOW | AUDITVNPATH1,
UIO_USERSPACE, uap->path, ctxp);
error = namei(ndp);
if (error != 0) {
goto out;
}
vp = ndp->ni_vp;
mp = vp->v_mount;
nameidone(ndp);
#if CONFIG_MACF
error = mac_mount_check_stat(ctxp, mp);
if (error != 0) {
vnode_put(vp);
goto out;
}
#endif
error = vfs_update_vfsstat(mp, ctxp, VFS_USER_EVENT);
if (error != 0) {
vnode_put(vp);
goto out;
}
sfsp = &__nameidata_statfs64->sfs;
vfs_get_statfs64(mp, sfsp);
if ((mp->mnt_kern_flag & MNTK_SYSTEMDATA) &&
(p->p_vfs_iopolicy & P_VFS_IOPOLICY_STATFS_NO_DATA_VOLUME)) {
/* This process does not want to see a seperate data volume mountpoint */
strlcpy(&sfsp->f_mntonname[0], "/", sizeof("/"));
}
error = copyout(sfsp, uap->buf, sizeof(*sfsp));
vnode_put(vp);
out:
kfree_type(typeof(*__nameidata_statfs64), __nameidata_statfs64);
return error;
}
/*
* Get file system statistics in 64-bit mode
*/
int
fstatfs64(__unused struct proc *p, struct fstatfs64_args *uap, __unused int32_t *retval)
{
struct vnode *vp;
struct mount *mp;
struct statfs64 sfs;
int error;
AUDIT_ARG(fd, uap->fd);
if ((error = file_vnode(uap->fd, &vp))) {
return error;
}
error = vnode_getwithref(vp);
if (error) {
file_drop(uap->fd);
return error;
}
AUDIT_ARG(vnpath_withref, vp, ARG_VNODE1);
mp = vp->v_mount;
if (!mp) {
error = EBADF;
goto out;
}
#if CONFIG_MACF
error = mac_mount_check_stat(vfs_context_current(), mp);
if (error != 0) {
goto out;
}
#endif
if ((error = vfs_update_vfsstat(mp, vfs_context_current(), VFS_USER_EVENT)) != 0) {
goto out;
}
vfs_get_statfs64(mp, &sfs);
if ((mp->mnt_kern_flag & MNTK_SYSTEMDATA) &&
(p->p_vfs_iopolicy & P_VFS_IOPOLICY_STATFS_NO_DATA_VOLUME)) {
/* This process does not want to see a seperate data volume mountpoint */
strlcpy(&sfs.f_mntonname[0], "/", sizeof("/"));
}
error = copyout(&sfs, uap->buf, sizeof(sfs));
out:
file_drop(uap->fd);
vnode_put(vp);
return error;
}
struct getfsstat_struct {
user_addr_t sfsp;
user_addr_t *mp;
int count;
int maxcount;
int flags;
int error;
};
static int
getfsstat_callback(mount_t mp, void * arg)
{
struct getfsstat_struct *fstp = (struct getfsstat_struct *)arg;
struct vfsstatfs *sp;
int error, my_size;
vfs_context_t ctx = vfs_context_current();
if (fstp->sfsp && fstp->count < fstp->maxcount) {
#if CONFIG_MACF
error = mac_mount_check_stat(ctx, mp);
if (error != 0) {
fstp->error = error;
return VFS_RETURNED_DONE;
}
#endif
sp = &mp->mnt_vfsstat;
/*
* If MNT_NOWAIT is specified, do not refresh the
* fsstat cache. MNT_WAIT/MNT_DWAIT overrides MNT_NOWAIT.
*/
if ((mp->mnt_lflag & MNT_LDEAD) ||
(((fstp->flags & MNT_NOWAIT) == 0 || (fstp->flags & (MNT_WAIT | MNT_DWAIT))) &&
(!(mp->mnt_lflag & MNT_LUNMOUNT)) &&
(error = vfs_update_vfsstat(mp, ctx, VFS_USER_EVENT)))) {
KAUTH_DEBUG("vfs_update_vfsstat returned %d", error);
return VFS_RETURNED;
}
/*
* Need to handle LP64 version of struct statfs
*/
error = munge_statfs(mp, sp, fstp->sfsp, &my_size, IS_64BIT_PROCESS(vfs_context_proc(ctx)), FALSE);
if (error) {
fstp->error = error;
return VFS_RETURNED_DONE;
}
fstp->sfsp += my_size;
if (fstp->mp) {
#if CONFIG_MACF
error = mac_mount_label_get(mp, *fstp->mp);
if (error) {
fstp->error = error;
return VFS_RETURNED_DONE;
}
#endif
fstp->mp++;
}
}
fstp->count++;
return VFS_RETURNED;
}
/*
* Get statistics on all filesystems.
*/
int
getfsstat(__unused proc_t p, struct getfsstat_args *uap, int *retval)
{
struct __mac_getfsstat_args muap;
muap.buf = uap->buf;
muap.bufsize = uap->bufsize;
muap.mac = USER_ADDR_NULL;
muap.macsize = 0;
muap.flags = uap->flags;
return __mac_getfsstat(p, &muap, retval);
}
/*
* __mac_getfsstat: Get MAC-related file system statistics
*
* Parameters: p (ignored)
* uap User argument descriptor (see below)
* retval Count of file system statistics (N stats)
*
* Indirect: uap->bufsize Buffer size
* uap->macsize MAC info size
* uap->buf Buffer where information will be returned
* uap->mac MAC info
* uap->flags File system flags
*
*
* Returns: 0 Success
* !0 Not success
*
*/
int
__mac_getfsstat(__unused proc_t p, struct __mac_getfsstat_args *uap, int *retval)
{
user_addr_t sfsp;
user_addr_t *mp;
size_t count, maxcount, bufsize, macsize;
struct getfsstat_struct fst;
if ((unsigned)uap->bufsize > INT_MAX || (unsigned)uap->macsize > INT_MAX) {
return EINVAL;
}
bufsize = (size_t) uap->bufsize;
macsize = (size_t) uap->macsize;
if (IS_64BIT_PROCESS(p)) {
maxcount = bufsize / sizeof(struct user64_statfs);
} else {
maxcount = bufsize / sizeof(struct user32_statfs);
}
sfsp = uap->buf;
count = 0;
mp = NULL;
#if CONFIG_MACF
if (uap->mac != USER_ADDR_NULL) {
u_int32_t *mp0;
int error;
unsigned int i;
count = (macsize / (IS_64BIT_PROCESS(p) ? 8 : 4));
if (count != maxcount) {
return EINVAL;
}
/* Copy in the array */
mp0 = kalloc_data(macsize, Z_WAITOK);
if (mp0 == NULL) {
return ENOMEM;
}
error = copyin(uap->mac, mp0, macsize);
if (error) {
kfree_data(mp0, macsize);
return error;
}
/* Normalize to an array of user_addr_t */
mp = kalloc_data(count * sizeof(user_addr_t), Z_WAITOK);
if (mp == NULL) {
kfree_data(mp0, macsize);
return ENOMEM;
}
for (i = 0; i < count; i++) {
if (IS_64BIT_PROCESS(p)) {
mp[i] = ((user_addr_t *)mp0)[i];
} else {
mp[i] = (user_addr_t)mp0[i];
}
}
kfree_data(mp0, macsize);
}
#endif
fst.sfsp = sfsp;
fst.mp = mp;
fst.flags = uap->flags;
fst.count = 0;
fst.error = 0;
fst.maxcount = (int)maxcount;
vfs_iterate(VFS_ITERATE_NOSKIP_UNMOUNT, getfsstat_callback, &fst);
if (mp) {
kfree_data(mp, count * sizeof(user_addr_t));
}
if (fst.error) {
KAUTH_DEBUG("ERROR - %s gets %d", p->p_comm, fst.error);
return fst.error;
}
if (fst.sfsp && fst.count > fst.maxcount) {
*retval = fst.maxcount;
} else {
*retval = fst.count;
}
return 0;
}
static int
getfsstat64_callback(mount_t mp, void * arg)
{
struct getfsstat_struct *fstp = (struct getfsstat_struct *)arg;
struct vfsstatfs *sp;
struct statfs64 sfs;
int error;
if (fstp->sfsp && fstp->count < fstp->maxcount) {
#if CONFIG_MACF
error = mac_mount_check_stat(vfs_context_current(), mp);
if (error != 0) {
fstp->error = error;
return VFS_RETURNED_DONE;
}
#endif
sp = &mp->mnt_vfsstat;
/*
* If MNT_NOWAIT is specified, do not refresh the fsstat
* cache. MNT_WAIT overrides MNT_NOWAIT.
*
* We treat MNT_DWAIT as MNT_WAIT for all instances of
* getfsstat, since the constants are out of the same
* namespace.
*/
if ((mp->mnt_lflag & MNT_LDEAD) ||
((((fstp->flags & MNT_NOWAIT) == 0) || (fstp->flags & (MNT_WAIT | MNT_DWAIT))) &&
(!(mp->mnt_lflag & MNT_LUNMOUNT)) &&
(error = vfs_update_vfsstat(mp, vfs_context_current(), VFS_USER_EVENT)))) {
KAUTH_DEBUG("vfs_update_vfsstat returned %d", error);
return VFS_RETURNED;
}
vfs_get_statfs64(mp, &sfs);
error = copyout(&sfs, fstp->sfsp, sizeof(sfs));
if (error) {
fstp->error = error;
return VFS_RETURNED_DONE;
}
fstp->sfsp += sizeof(sfs);
}
fstp->count++;
return VFS_RETURNED;
}
/*
* Get statistics on all file systems in 64 bit mode.
*/
int
getfsstat64(__unused proc_t p, struct getfsstat64_args *uap, int *retval)
{
user_addr_t sfsp;
int count, maxcount;
struct getfsstat_struct fst;
maxcount = uap->bufsize / sizeof(struct statfs64);
sfsp = uap->buf;
count = 0;
fst.sfsp = sfsp;
fst.flags = uap->flags;
fst.count = 0;
fst.error = 0;
fst.maxcount = maxcount;
vfs_iterate(VFS_ITERATE_NOSKIP_UNMOUNT, getfsstat64_callback, &fst);
if (fst.error) {
KAUTH_DEBUG("ERROR - %s gets %d", p->p_comm, fst.error);
return fst.error;
}
if (fst.sfsp && fst.count > fst.maxcount) {
*retval = fst.maxcount;
} else {
*retval = fst.count;
}
return 0;
}
/*
* gets the associated vnode with the file descriptor passed.
* as input
*
* INPUT
* ctx - vfs context of caller
* fd - file descriptor for which vnode is required.
* vpp - Pointer to pointer to vnode to be returned.
*
* The vnode is returned with an iocount so any vnode obtained
* by this call needs a vnode_put
*
*/
int
vnode_getfromfd(vfs_context_t ctx, int fd, vnode_t *vpp)
{
int error;
vnode_t vp;
struct fileproc *fp;
proc_t p = vfs_context_proc(ctx);
*vpp = NULLVP;
error = fp_getfvp(p, fd, &fp, &vp);
if (error) {
return error;
}
error = vnode_getwithref(vp);
if (error) {
(void)fp_drop(p, fd, fp, 0);
return error;
}
(void)fp_drop(p, fd, fp, 0);
*vpp = vp;
return error;
}
/*
* Wrapper function around namei to start lookup from a directory
* specified by a file descriptor ni_dirfd.
*
* In addition to all the errors returned by namei, this call can
* return ENOTDIR if the file descriptor does not refer to a directory.
* and EBADF if the file descriptor is not valid.
*/
int
nameiat(struct nameidata *ndp, int dirfd)
{
if ((dirfd != AT_FDCWD) &&
!(ndp->ni_flag & NAMEI_CONTLOOKUP) &&
!(ndp->ni_cnd.cn_flags & USEDVP)) {
int error = 0;
char c;
if (UIO_SEG_IS_USER_SPACE(ndp->ni_segflg)) {
error = copyin(ndp->ni_dirp, &c, sizeof(char));
if (error) {
return error;
}
} else {
c = *((char *)(ndp->ni_dirp));
}
if (c != '/') {
vnode_t dvp_at;
error = vnode_getfromfd(ndp->ni_cnd.cn_context, dirfd,
&dvp_at);
if (error) {
return error;
}
if (vnode_vtype(dvp_at) != VDIR) {
vnode_put(dvp_at);
return ENOTDIR;
}
ndp->ni_dvp = dvp_at;
ndp->ni_cnd.cn_flags |= USEDVP;
error = namei(ndp);
ndp->ni_cnd.cn_flags &= ~USEDVP;
vnode_put(dvp_at);
return error;
}
}
return namei(ndp);
}
/*
* Change current working directory to a given file descriptor.
*/
/* ARGSUSED */
int
fchdir(proc_t p, vfs_context_t ctx, int fd, bool per_thread)
{
vnode_t vp;
vnode_t tdp;
vnode_t tvp;
struct mount *mp;
int error, should_put = 1;
AUDIT_ARG(fd, fd);
if (per_thread && fd == -1) {
/*
* Switching back from per-thread to per process CWD; verify we
* in fact have one before proceeding. The only success case
* for this code path is to return 0 preemptively after zapping
* the thread structure contents.
*/
thread_t th = vfs_context_thread(ctx);
if (th) {
uthread_t uth = get_bsdthread_info(th);
tvp = uth->uu_cdir;
uth->uu_cdir = NULLVP;
if (tvp != NULLVP) {
vnode_rele(tvp);
return 0;
}
}
return EBADF;
}
if ((error = file_vnode(fd, &vp))) {
return error;
}
if ((error = vnode_getwithref(vp))) {
file_drop(fd);
return error;
}
AUDIT_ARG(vnpath, vp, ARG_VNODE1);
if (vp->v_type != VDIR) {
error = ENOTDIR;
goto out;
}
#if CONFIG_MACF
error = mac_vnode_check_chdir(ctx, vp);
if (error) {
goto out;
}
#endif
error = vnode_authorize(vp, NULL, KAUTH_VNODE_SEARCH, ctx);
if (error) {
goto out;
}
while (!error && (mp = vp->v_mountedhere) != NULL) {
if (vfs_busy(mp, LK_NOWAIT)) {
error = EACCES;
goto out;
}
error = VFS_ROOT(mp, &tdp, ctx);
vfs_unbusy(mp);
if (error) {
break;
}
vnode_put(vp);
vp = tdp;
}
if (error) {
goto out;
}
if ((error = vnode_ref(vp))) {
goto out;
}
vnode_put(vp);
should_put = 0;
if (per_thread) {
thread_t th = vfs_context_thread(ctx);
if (th) {
uthread_t uth = get_bsdthread_info(th);
tvp = uth->uu_cdir;
uth->uu_cdir = vp;
OSBitOrAtomic(P_THCWD, &p->p_flag);
} else {
vnode_rele(vp);
error = ENOENT;
goto out;
}
} else {
proc_dirs_lock_exclusive(p);
proc_fdlock(p);
tvp = p->p_fd.fd_cdir;
p->p_fd.fd_cdir = vp;
proc_fdunlock(p);
proc_dirs_unlock_exclusive(p);
}
if (tvp) {
vnode_rele(tvp);
}
out:
if (should_put) {
vnode_put(vp);
}
file_drop(fd);
return error;
}
int
sys_fchdir(proc_t p, struct fchdir_args *uap, __unused int32_t *retval)
{
return fchdir(p, vfs_context_current(), uap->fd, false);
}
int
__pthread_fchdir(proc_t p, struct __pthread_fchdir_args *uap, __unused int32_t *retval)
{
return fchdir(p, vfs_context_current(), uap->fd, true);
}
/*
* Change current working directory (".").
*
* Returns: 0 Success
* change_dir:ENOTDIR
* change_dir:???
* vnode_ref:ENOENT No such file or directory
*/
/* ARGSUSED */
int
chdir_internal(proc_t p, vfs_context_t ctx, struct nameidata *ndp, int per_thread)
{
int error;
vnode_t tvp;
error = change_dir(ndp, ctx);
if (error) {
return error;
}
if ((error = vnode_ref(ndp->ni_vp))) {
vnode_put(ndp->ni_vp);
return error;
}
/*
* drop the iocount we picked up in change_dir
*/
vnode_put(ndp->ni_vp);
if (per_thread) {
thread_t th = vfs_context_thread(ctx);
if (th) {
uthread_t uth = get_bsdthread_info(th);
tvp = uth->uu_cdir;
uth->uu_cdir = ndp->ni_vp;
OSBitOrAtomic(P_THCWD, &p->p_flag);
} else {
vnode_rele(ndp->ni_vp);
return ENOENT;
}
} else {
proc_dirs_lock_exclusive(p);
proc_fdlock(p);
tvp = p->p_fd.fd_cdir;
p->p_fd.fd_cdir = ndp->ni_vp;
proc_fdunlock(p);
proc_dirs_unlock_exclusive(p);
}
if (tvp) {
vnode_rele(tvp);
}
return 0;
}
/*
* Change current working directory (".").
*
* Returns: 0 Success
* chdir_internal:ENOTDIR
* chdir_internal:ENOENT No such file or directory
* chdir_internal:???
*/
/* ARGSUSED */
static int
common_chdir(proc_t p, struct chdir_args *uap, int per_thread)
{
struct nameidata nd;
vfs_context_t ctx = vfs_context_current();
NDINIT(&nd, LOOKUP, OP_CHDIR, FOLLOW | AUDITVNPATH1,
UIO_USERSPACE, uap->path, ctx);
return chdir_internal(p, ctx, &nd, per_thread);
}
/*
* chdir
*
* Change current working directory (".") for the entire process
*
* Parameters: p Process requesting the call
* uap User argument descriptor (see below)
* retval (ignored)
*
* Indirect parameters: uap->path Directory path
*
* Returns: 0 Success
* common_chdir: ENOTDIR
* common_chdir: ENOENT No such file or directory
* common_chdir: ???
*
*/
int
sys_chdir(proc_t p, struct chdir_args *uap, __unused int32_t *retval)
{
return common_chdir(p, (void *)uap, 0);
}
/*
* __pthread_chdir
*
* Change current working directory (".") for a single thread
*
* Parameters: p Process requesting the call
* uap User argument descriptor (see below)
* retval (ignored)
*
* Indirect parameters: uap->path Directory path
*
* Returns: 0 Success
* common_chdir: ENOTDIR
* common_chdir: ENOENT No such file or directory
* common_chdir: ???
*
*/
int
__pthread_chdir(proc_t p, struct __pthread_chdir_args *uap, __unused int32_t *retval)
{
return common_chdir(p, (void *)uap, 1);
}
/*
* Change notion of root (``/'') directory.
*/
/* ARGSUSED */
int
chroot(proc_t p, struct chroot_args *uap, __unused int32_t *retval)
{
struct filedesc *fdp = &p->p_fd;
int error;
struct nameidata nd;
vnode_t tvp;
vfs_context_t ctx = vfs_context_current();
if ((error = suser(kauth_cred_get(), &p->p_acflag))) {
return error;
}
NDINIT(&nd, LOOKUP, OP_CHROOT, FOLLOW | AUDITVNPATH1,
UIO_USERSPACE, uap->path, ctx);
error = change_dir(&nd, ctx);
if (error) {
return error;
}
#if CONFIG_MACF
error = mac_vnode_check_chroot(ctx, nd.ni_vp,
&nd.ni_cnd);
if (error) {
vnode_put(nd.ni_vp);
return error;
}
#endif
if ((error = vnode_ref(nd.ni_vp))) {
vnode_put(nd.ni_vp);
return error;
}
vnode_put(nd.ni_vp);
/*
* This lock provides the guarantee that as long as you hold the lock
* fdp->fd_rdir has a usecount on it. This is used to take an iocount
* on a referenced vnode in namei when determining the rootvnode for
* a process.
*/
/* needed for synchronization with lookup */
proc_dirs_lock_exclusive(p);
/* needed for setting the flag and other activities on the fd itself */
proc_fdlock(p);
tvp = fdp->fd_rdir;
fdp->fd_rdir = nd.ni_vp;
fdt_flag_set(fdp, FD_CHROOT);
proc_fdunlock(p);
proc_dirs_unlock_exclusive(p);
if (tvp != NULL) {
vnode_rele(tvp);
}
return 0;
}
#define PATHSTATICBUFLEN 256
#define PIVOT_ROOT_ENTITLEMENT \
"com.apple.private.vfs.pivot-root"
#if defined(XNU_TARGET_OS_OSX)
int
pivot_root(proc_t p, struct pivot_root_args *uap, __unused int *retval)
{
int error;
char new_rootfs_path_before[PATHSTATICBUFLEN] = {0};
char old_rootfs_path_after[PATHSTATICBUFLEN] = {0};
char *new_rootfs_path_before_buf = NULL;
char *old_rootfs_path_after_buf = NULL;
char *incoming = NULL;
char *outgoing = NULL;
vnode_t incoming_rootvp = NULLVP;
size_t bytes_copied;
/*
* XXX : Additional restrictions needed
* - perhaps callable only once.
*/
if ((error = suser(kauth_cred_get(), &p->p_acflag))) {
return error;
}
/*
* pivot_root can be executed by launchd only.
* Enforce entitlement.
*/
if ((proc_getpid(p) != 1) || !IOCurrentTaskHasEntitlement(PIVOT_ROOT_ENTITLEMENT)) {
return EPERM;
}
error = copyinstr(uap->new_rootfs_path_before, &new_rootfs_path_before[0], PATHSTATICBUFLEN, &bytes_copied);
if (error == ENAMETOOLONG) {
new_rootfs_path_before_buf = zalloc_flags(ZV_NAMEI, Z_WAITOK);
error = copyinstr(uap->new_rootfs_path_before, new_rootfs_path_before_buf, MAXPATHLEN, &bytes_copied);
}
if (error) {
goto out;
}
error = copyinstr(uap->old_rootfs_path_after, &old_rootfs_path_after[0], PATHSTATICBUFLEN, &bytes_copied);
if (error == ENAMETOOLONG) {
old_rootfs_path_after_buf = zalloc_flags(ZV_NAMEI, Z_WAITOK);
error = copyinstr(uap->old_rootfs_path_after, old_rootfs_path_after_buf, MAXPATHLEN, &bytes_copied);
}
if (error) {
goto out;
}
if (new_rootfs_path_before_buf) {
incoming = new_rootfs_path_before_buf;
} else {
incoming = &new_rootfs_path_before[0];
}
if (old_rootfs_path_after_buf) {
outgoing = old_rootfs_path_after_buf;
} else {
outgoing = &old_rootfs_path_after[0];
}
/*
* The proposed incoming FS MUST be authenticated (i.e. not a chunklist DMG).
* Userland is not allowed to pivot to an image.
*/
error = vnode_lookup(incoming, 0, &incoming_rootvp, vfs_context_kernel());
if (error) {
goto out;
}
error = VNOP_IOCTL(incoming_rootvp, FSIOC_KERNEL_ROOTAUTH, NULL, 0, vfs_context_kernel());
if (error) {
goto out;
}
error = vfs_switch_root(incoming, outgoing, VFSSR_VIRTUALDEV_PROHIBITED);
out:
if (incoming_rootvp != NULLVP) {
vnode_put(incoming_rootvp);
incoming_rootvp = NULLVP;
}
if (old_rootfs_path_after_buf) {
zfree(ZV_NAMEI, old_rootfs_path_after_buf);
}
if (new_rootfs_path_before_buf) {
zfree(ZV_NAMEI, new_rootfs_path_before_buf);
}
return error;
}
#else
int
pivot_root(proc_t p, __unused struct pivot_root_args *uap, int *retval)
{
return nosys(p, NULL, retval);
}
#endif /* XNU_TARGET_OS_OSX */
/*
* Common routine for chroot and chdir.
*
* Returns: 0 Success
* ENOTDIR Not a directory
* namei:??? [anything namei can return]
* vnode_authorize:??? [anything vnode_authorize can return]
*/
static int
change_dir(struct nameidata *ndp, vfs_context_t ctx)
{
vnode_t vp;
int error;
if ((error = namei(ndp))) {
return error;
}
nameidone(ndp);
vp = ndp->ni_vp;
if (vp->v_type != VDIR) {
vnode_put(vp);
return ENOTDIR;
}
#if CONFIG_MACF
error = mac_vnode_check_chdir(ctx, vp);
if (error) {
vnode_put(vp);
return error;
}
#endif
error = vnode_authorize(vp, NULL, KAUTH_VNODE_SEARCH, ctx);
if (error) {
vnode_put(vp);
return error;
}
return error;
}
/*
* Free the vnode data (for directories) associated with the file glob.
*/
struct fd_vn_data *
fg_vn_data_alloc(void)
{
struct fd_vn_data *fvdata;
/* Allocate per fd vnode data */
fvdata = kalloc_type(struct fd_vn_data, Z_WAITOK | Z_ZERO);
lck_mtx_init(&fvdata->fv_lock, &fd_vn_lck_grp, &fd_vn_lck_attr);
return fvdata;
}
/*
* Free the vnode data (for directories) associated with the file glob.
*/
void
fg_vn_data_free(void *fgvndata)
{
struct fd_vn_data *fvdata = (struct fd_vn_data *)fgvndata;
kfree_data(fvdata->fv_buf, fvdata->fv_bufallocsiz);
lck_mtx_destroy(&fvdata->fv_lock, &fd_vn_lck_grp);
kfree_type(struct fd_vn_data, fvdata);
}
/*
* Check permissions, allocate an open file structure,
* and call the device open routine if any.
*
* Returns: 0 Success
* EINVAL
* EINTR
* falloc:ENFILE
* falloc:EMFILE
* falloc:ENOMEM
* vn_open_auth:???
* dupfdopen:???
* VNOP_ADVLOCK:???
* vnode_setsize:???
*
* XXX Need to implement uid, gid
*/
int
open1(vfs_context_t ctx, struct nameidata *ndp, int uflags,
struct vnode_attr *vap, fp_initfn_t fp_init, void *initarg, int32_t *retval, int authfd)
{
proc_t p = vfs_context_proc(ctx);
kauth_cred_t p_cred = current_cached_proc_cred(PROC_NULL);
uthread_t uu = get_bsdthread_info(vfs_context_thread(ctx));
struct fileproc *fp;
vnode_t vp;
int flags, oflags, amode;
int type, indx, error;
struct vfs_context context;
vnode_t authvp = NULLVP;
oflags = uflags;
amode = oflags & O_ACCMODE;
/*
* Because O_RDONLY is 0, it is not possible to distinguish between
* O_EXEC | O_RDONLY and O_EXEC, therefore FEXEC/FSEARCH can't be set together
* with FREAD/FWRITE.
*/
if ((amode == O_ACCMODE) || (amode && (oflags & O_EXEC))) {
return EINVAL;
}
flags = FFLAGS(uflags);
CLR(flags, FENCRYPTED);
CLR(flags, FUNENCRYPTED);
AUDIT_ARG(fflags, oflags);
AUDIT_ARG(mode, vap->va_mode);
if ((error = falloc_withinit(p, p_cred, ctx, &fp, &indx, fp_init, initarg)) != 0) {
return error;
}
if (flags & O_CLOEXEC) {
fp->fp_flags |= FP_CLOEXEC;
}
if (flags & O_CLOFORK) {
fp->fp_flags |= FP_CLOFORK;
}
/* setup state to recognize when fdesc_open was called */
uu->uu_dupfd = -1;
/*
* Disable read/write access if file is opened with O_EVTONLY and
* the process has requested to deny read/write access.
*/
if ((flags & O_EVTONLY) && proc_disallow_rw_for_o_evtonly(p)) {
flags &= ~(FREAD | FWRITE);
}
if (authfd != AUTH_OPEN_NOAUTHFD) {
error = vnode_getfromfd(ctx, authfd, &authvp);
if (error) {
fp_free(p, indx, fp);
return error;
}
}
if ((error = vn_open_auth(ndp, &flags, vap, authvp))) {
if (authvp != NULLVP) {
vnode_put(authvp);
}
if ((error == ENODEV || error == ENXIO) && (uu->uu_dupfd >= 0)) {
if ((error = dupfdopen(p, indx, uu->uu_dupfd, flags, error)) == 0) {
*retval = indx;
return 0;
}
}
if (error == ERESTART) {
error = EINTR;
}
fp_free(p, indx, fp);
return error;
}
if (authvp != NULLVP) {
vnode_put(authvp);
}
uu->uu_dupfd = 0;
vp = ndp->ni_vp;
fp->fp_glob->fg_flag = flags & (FMASK | O_EVTONLY | FENCRYPTED | FUNENCRYPTED);
fp->fp_glob->fg_ops = &vnops;
fp_set_data(fp, vp);
#if CONFIG_FILE_LEASES
/*
* If we are creating a file or open with truncate, we need to break the
* lease if there is a read lease placed on the parent dir.
*/
if ((vnode_vtype(vp) == VREG) && (flags & (O_CREAT | O_TRUNC))) {
vnode_breakdirlease(vp, true, oflags);
}
/* Now check if there is a lease placed on the file itself. */
error = vnode_breaklease(vp, oflags, ctx);
if (error) {
goto bad;
}
#endif /* CONFIG_FILE_LEASES */
if (flags & (O_EXLOCK | O_SHLOCK)) {
struct flock lf = {
.l_whence = SEEK_SET,
};
if (flags & O_EXLOCK) {
lf.l_type = F_WRLCK;
} else {
lf.l_type = F_RDLCK;
}
type = F_FLOCK;
if ((flags & FNONBLOCK) == 0) {
type |= F_WAIT;
}
#if CONFIG_MACF
error = mac_file_check_lock(vfs_context_ucred(ctx), fp->fp_glob,
F_SETLK, &lf);
if (error) {
goto bad;
}
#endif
if ((error = VNOP_ADVLOCK(vp, (caddr_t)fp->fp_glob, F_SETLK, &lf, type, ctx, NULL))) {
goto bad;
}
fp->fp_glob->fg_flag |= FWASLOCKED;
}
/* try to truncate by setting the size attribute */
if ((flags & O_TRUNC) && ((error = vnode_setsize(vp, (off_t)0, 0, ctx)) != 0)) {
goto bad;
}
/*
* For directories we hold some additional information in the fd.
*/
if (vnode_vtype(vp) == VDIR) {
fp->fp_glob->fg_vn_data = fg_vn_data_alloc();
} else {
fp->fp_glob->fg_vn_data = NULL;
}
#if CONFIG_SECLUDED_MEMORY
if (secluded_for_filecache && vnode_vtype(vp) == VREG) {
memory_object_control_t moc;
const char *v_name;
moc = ubc_getobject(vp, UBC_FLAGS_NONE);
if (moc == MEMORY_OBJECT_CONTROL_NULL) {
/* nothing to do... */
} else if (fp->fp_glob->fg_flag & FWRITE) {
/* writable -> no longer eligible for secluded pages */
memory_object_mark_eligible_for_secluded(moc,
FALSE);
} else if (secluded_for_filecache == SECLUDED_FILECACHE_APPS) {
char pathname[32] = { 0, };
size_t copied;
/* XXX FBDP: better way to detect /Applications/ ? */
if (UIO_SEG_IS_USER_SPACE(ndp->ni_segflg)) {
(void)copyinstr(ndp->ni_dirp,
pathname,
sizeof(pathname),
&copied);
} else {
copystr(CAST_DOWN(void *, ndp->ni_dirp),
pathname,
sizeof(pathname),
&copied);
}
pathname[sizeof(pathname) - 1] = '\0';
if (strncmp(pathname,
"/Applications/",
strlen("/Applications/")) == 0 &&
strncmp(pathname,
"/Applications/Camera.app/",
strlen("/Applications/Camera.app/")) != 0) {
/*
* not writable
* AND from "/Applications/"
* AND not from "/Applications/Camera.app/"
* ==> eligible for secluded
*/
memory_object_mark_eligible_for_secluded(moc,
TRUE);
}
} else if (secluded_for_filecache == SECLUDED_FILECACHE_RDONLY &&
(v_name = vnode_getname(vp))) {
size_t len = strlen(v_name);
if (!strncmp(v_name, "dyld", len) ||
!strncmp(v_name, "launchd", len) ||
!strncmp(v_name, "Camera", len) ||
!strncmp(v_name, "SpringBoard", len) ||
!strncmp(v_name, "backboardd", len)) {
/*
* This file matters when launching Camera:
* do not store its contents in the secluded
* pool that will be drained on Camera launch.
*/
memory_object_mark_eligible_for_secluded(moc,
FALSE);
} else if (!strncmp(v_name, "mediaserverd", len)) {
memory_object_mark_eligible_for_secluded(moc,
FALSE);
memory_object_mark_for_realtime(moc,
true);
} else if (!strncmp(v_name, "bluetoothd", len)) {
/*
* bluetoothd might be needed for realtime audio
* playback.
*/
memory_object_mark_eligible_for_secluded(moc,
FALSE);
memory_object_mark_for_realtime(moc,
true);
} else {
char pathname[64] = { 0, };
size_t copied;
if (UIO_SEG_IS_USER_SPACE(ndp->ni_segflg)) {
(void)copyinstr(ndp->ni_dirp,
pathname,
sizeof(pathname),
&copied);
} else {
copystr(CAST_DOWN(void *, ndp->ni_dirp),
pathname,
sizeof(pathname),
&copied);
}
pathname[sizeof(pathname) - 1] = '\0';
if (strncmp(pathname,
"/Library/Audio/Plug-Ins/",
strlen("/Library/Audio/Plug-Ins/")) == 0 ||
strncmp(pathname,
"/System/Library/Audio/Plug-Ins/",
strlen("/System/Library/Audio/Plug-Ins/")) == 0) {
/*
* This may be an audio plugin required
* for realtime playback.
* ==> NOT eligible for secluded.
*/
memory_object_mark_eligible_for_secluded(moc,
FALSE);
memory_object_mark_for_realtime(moc,
true);
}
}
vnode_putname(v_name);
}
}
#endif /* CONFIG_SECLUDED_MEMORY */
vnode_put(vp);
/*
* The first terminal open (without a O_NOCTTY) by a session leader
* results in it being set as the controlling terminal.
*/
if (vnode_istty(vp) && !(p->p_flag & P_CONTROLT) &&
!(flags & O_NOCTTY)) {
int tmp = 0;
(void)(*fp->fp_glob->fg_ops->fo_ioctl)(fp, (int)TIOCSCTTY,
(caddr_t)&tmp, ctx);
}
proc_fdlock(p);
procfdtbl_releasefd(p, indx, NULL);
fp_drop(p, indx, fp, 1);
proc_fdunlock(p);
*retval = indx;
return 0;
bad:
context = *vfs_context_current();
context.vc_ucred = fp->fp_glob->fg_cred;
if ((fp->fp_glob->fg_flag & FWASLOCKED) &&
(FILEGLOB_DTYPE(fp->fp_glob) == DTYPE_VNODE)) {
struct flock lf = {
.l_whence = SEEK_SET,
.l_type = F_UNLCK,
};
(void)VNOP_ADVLOCK(
vp, (caddr_t)fp->fp_glob, F_UNLCK, &lf, F_FLOCK, ctx, NULL);
}
vn_close(vp, fp->fp_glob->fg_flag, &context);
vnode_put(vp);
fp_free(p, indx, fp);
return error;
}
/*
* While most of the *at syscall handlers can call nameiat() which
* is a wrapper around namei, the use of namei and initialisation
* of nameidata are far removed and in different functions - namei
* gets called in vn_open_auth for open1. So we'll just do here what
* nameiat() does.
*/
static int
open1at(vfs_context_t ctx, struct nameidata *ndp, int uflags,
struct vnode_attr *vap, fp_initfn_t fp_init, void *initarg, int32_t *retval,
int dirfd, int authfd)
{
if ((dirfd != AT_FDCWD) && !(ndp->ni_cnd.cn_flags & USEDVP)) {
int error;
char c;
if (UIO_SEG_IS_USER_SPACE(ndp->ni_segflg)) {
error = copyin(ndp->ni_dirp, &c, sizeof(char));
if (error) {
return error;
}
} else {
c = *((char *)(ndp->ni_dirp));
}
if (c != '/') {
vnode_t dvp_at;
error = vnode_getfromfd(ndp->ni_cnd.cn_context, dirfd,
&dvp_at);
if (error) {
return error;
}
if (vnode_vtype(dvp_at) != VDIR) {
vnode_put(dvp_at);
return ENOTDIR;
}
ndp->ni_dvp = dvp_at;
ndp->ni_cnd.cn_flags |= USEDVP;
error = open1(ctx, ndp, uflags, vap, fp_init, initarg,
retval, authfd);
vnode_put(dvp_at);
return error;
}
}
return open1(ctx, ndp, uflags, vap, fp_init, initarg, retval, authfd);
}
/*
* open_extended: open a file given a path name; with extended argument list (including extended security (ACL)).
*
* Parameters: p Process requesting the open
* uap User argument descriptor (see below)
* retval Pointer to an area to receive the
* return calue from the system call
*
* Indirect: uap->path Path to open (same as 'open')
* uap->flags Flags to open (same as 'open'
* uap->uid UID to set, if creating
* uap->gid GID to set, if creating
* uap->mode File mode, if creating (same as 'open')
* uap->xsecurity ACL to set, if creating
*
* Returns: 0 Success
* !0 errno value
*
* Notes: The kauth_filesec_t in 'va', if any, is in host byte order.
*
* XXX: We should enummerate the possible errno values here, and where
* in the code they originated.
*/
int
open_extended(proc_t p, struct open_extended_args *uap, int32_t *retval)
{
int ciferror;
kauth_filesec_t xsecdst;
struct vnode_attr va;
struct nameidata nd;
int cmode;
AUDIT_ARG(owner, uap->uid, uap->gid);
xsecdst = NULL;
if ((uap->xsecurity != USER_ADDR_NULL) &&
((ciferror = kauth_copyinfilesec(uap->xsecurity, &xsecdst)) != 0)) {
return ciferror;
}
VATTR_INIT(&va);
cmode = ((uap->mode & ~p->p_fd.fd_cmask) & ALLPERMS) & ~S_ISTXT;
VATTR_SET(&va, va_mode, cmode & ACCESSPERMS);
if (uap->uid != KAUTH_UID_NONE) {
VATTR_SET(&va, va_uid, uap->uid);
}
if (uap->gid != KAUTH_GID_NONE) {
VATTR_SET(&va, va_gid, uap->gid);
}
if (xsecdst != NULL) {
VATTR_SET(&va, va_acl, &xsecdst->fsec_acl);
va.va_vaflags |= VA_FILESEC_ACL;
}
NDINIT(&nd, LOOKUP, OP_OPEN, FOLLOW | AUDITVNPATH1, UIO_USERSPACE,
uap->path, vfs_context_current());
ciferror = open1(vfs_context_current(), &nd, uap->flags, &va,
NULL, NULL, retval, AUTH_OPEN_NOAUTHFD);
if (xsecdst != NULL) {
kauth_filesec_free(xsecdst);
}
return ciferror;
}
/*
* Go through the data-protected atomically controlled open (2)
*
* int open_dprotected_np(user_addr_t path, int flags, int class, int dpflags, int mode)
*/
static int
openat_dprotected_internal(vfs_context_t ctx, user_addr_t path, int flags, int mode,
int class, int dpflags, int fd, int authfd, enum uio_seg segflg, int *retval)
{
/*
* Follow the same path as normal open(2)
* Look up the item if it exists, and acquire the vnode.
*/
struct vnode_attr va;
struct nameidata nd;
int cmode;
int error;
struct filedesc *fdp = &vfs_context_proc(ctx)->p_fd;
VATTR_INIT(&va);
/* Mask off all but regular access permissions */
cmode = ((mode & ~fdp->fd_cmask) & ALLPERMS) & ~S_ISTXT;
VATTR_SET(&va, va_mode, cmode & ACCESSPERMS);
NDINIT(&nd, LOOKUP, OP_OPEN, FOLLOW | AUDITVNPATH1, segflg,
path, ctx);
/*
* Initialize the extra fields in vnode_attr to pass down our
* extra fields.
* 1. target cprotect class.
* 2. set a flag to mark it as requiring open-raw-encrypted semantics.
*/
if (flags & O_CREAT) {
/* lower level kernel code validates that the class is valid before applying it. */
if (class != PROTECTION_CLASS_DEFAULT) {
/*
* PROTECTION_CLASS_DEFAULT implies that we make the class for this
* file behave the same as open (2)
*/
VATTR_SET(&va, va_dataprotect_class, class);
}
}
if (dpflags & (O_DP_GETRAWENCRYPTED | O_DP_GETRAWUNENCRYPTED | O_DP_AUTHENTICATE)) {
if (flags & (O_RDWR | O_WRONLY)) {
/*
* Not allowed to write raw encrypted bytes or when opening authenticated.
*/
return EINVAL;
}
if (dpflags & O_DP_GETRAWENCRYPTED) {
VATTR_SET(&va, va_dataprotect_flags, VA_DP_RAWENCRYPTED);
}
if (dpflags & O_DP_GETRAWUNENCRYPTED) {
VATTR_SET(&va, va_dataprotect_flags, VA_DP_RAWUNENCRYPTED);
}
if (dpflags & O_DP_AUTHENTICATE) {
VATTR_SET(&va, va_dataprotect_flags, VA_DP_AUTHENTICATE);
}
}
error = open1at(vfs_context_current(), &nd, flags, &va,
NULL, NULL, retval, fd, authfd);
return error;
}
int
openat_dprotected_np(__unused proc_t p, struct openat_dprotected_np_args *uap, int32_t *retval)
{
if ((uap->dpflags & O_DP_AUTHENTICATE) && (uap->flags & O_CREAT)) {
return EINVAL;
}
return openat_dprotected_internal(vfs_context_current(), uap->path, uap->flags, uap->mode,
uap->class, uap->dpflags, uap->fd, uap->authfd, UIO_USERSPACE, retval);
}
int
open_dprotected_np(__unused proc_t p, struct open_dprotected_np_args *uap, int32_t *retval)
{
if (uap->dpflags & O_DP_AUTHENTICATE) {
return EINVAL;
}
return openat_dprotected_internal(vfs_context_current(), uap->path, uap->flags, uap->mode,
uap->class, uap->dpflags, AT_FDCWD, AUTH_OPEN_NOAUTHFD, UIO_USERSPACE, retval);
}
static int
openat_internal(vfs_context_t ctx, user_addr_t path, int flags, int mode,
int fd, enum uio_seg segflg, int *retval)
{
struct filedesc *fdp = &vfs_context_proc(ctx)->p_fd;
struct {
struct vnode_attr va;
struct nameidata nd;
} *__open_data;
struct vnode_attr *vap;
struct nameidata *ndp;
int cmode;
int error;
__open_data = kalloc_type(typeof(*__open_data), Z_WAITOK);
vap = &__open_data->va;
ndp = &__open_data->nd;
VATTR_INIT(vap);
/* Mask off all but regular access permissions */
cmode = ((mode & ~fdp->fd_cmask) & ALLPERMS) & ~S_ISTXT;
VATTR_SET(vap, va_mode, cmode & ACCESSPERMS);
NDINIT(ndp, LOOKUP, OP_OPEN, FOLLOW | AUDITVNPATH1,
segflg, path, ctx);
error = open1at(ctx, ndp, flags, vap, NULL, NULL, retval, fd, AUTH_OPEN_NOAUTHFD);
kfree_type(typeof(*__open_data), __open_data);
return error;
}
int
open(proc_t p, struct open_args *uap, int32_t *retval)
{
__pthread_testcancel(1);
return open_nocancel(p, (struct open_nocancel_args *)uap, retval);
}
int
open_nocancel(__unused proc_t p, struct open_nocancel_args *uap,
int32_t *retval)
{
return openat_internal(vfs_context_current(), uap->path, uap->flags,
uap->mode, AT_FDCWD, UIO_USERSPACE, retval);
}
int
openat_nocancel(__unused proc_t p, struct openat_nocancel_args *uap,
int32_t *retval)
{
return openat_internal(vfs_context_current(), uap->path, uap->flags,
uap->mode, uap->fd, UIO_USERSPACE, retval);
}
int
openat(proc_t p, struct openat_args *uap, int32_t *retval)
{
__pthread_testcancel(1);
return openat_nocancel(p, (struct openat_nocancel_args *)uap, retval);
}
#define OPEN_BY_ID_ENTITLEMENT "com.apple.private.vfs.open-by-id"
static boolean_t
vfs_context_can_open_by_id(vfs_context_t ctx)
{
if (csproc_get_platform_binary(vfs_context_proc(ctx))) {
return TRUE;
}
return IOTaskHasEntitlement(vfs_context_task(ctx),
OPEN_BY_ID_ENTITLEMENT);
}
/*
* openbyid_np: open a file given a file system id and a file system object id
* the hfs file system object id is an fsobj_id_t {uint32, uint32}
* file systems that don't support object ids it is a node id (uint64_t).
*
* Parameters: p Process requesting the open
* uap User argument descriptor (see below)
* retval Pointer to an area to receive the
* return calue from the system call
*
* Indirect: uap->path Path to open (same as 'open')
*
* uap->fsid id of target file system
* uap->objid id of target file system object
* uap->flags Flags to open (same as 'open')
*
* Returns: 0 Success
* !0 errno value
*
*
* XXX: We should enummerate the possible errno values here, and where
* in the code they originated.
*/
int
openbyid_np(__unused proc_t p, struct openbyid_np_args *uap, int *retval)
{
fsid_t fsid;
uint64_t objid;
int error;
char *buf = NULL;
int buflen = MAXPATHLEN;
int pathlen = 0;
vfs_context_t ctx = vfs_context_current();
if (!vfs_context_can_open_by_id(ctx)) {
return EPERM;
}
if ((error = copyin(uap->fsid, (caddr_t)&fsid, sizeof(fsid)))) {
return error;
}
/*uap->obj is an fsobj_id_t defined as struct {uint32_t, uint32_t} */
if ((error = copyin(uap->objid, (caddr_t)&objid, sizeof(uint64_t)))) {
return error;
}
AUDIT_ARG(value32, fsid.val[0]);
AUDIT_ARG(value64, objid);
/*resolve path from fsis, objid*/
do {
buf = kalloc_data(buflen + 1, Z_WAITOK);
if (buf == NULL) {
return ENOMEM;
}
error = fsgetpath_internal( ctx, fsid.val[0], objid, buflen,
buf, FSOPT_ISREALFSID, &pathlen);
if (error) {
kfree_data(buf, buflen + 1);
buf = NULL;
}
} while (error == ENOSPC && (buflen += MAXPATHLEN));
if (error) {
return error;
}
buf[pathlen] = 0;
error = openat_internal(
ctx, (user_addr_t)buf, uap->oflags, 0, AT_FDCWD, UIO_SYSSPACE, retval);
kfree_data(buf, buflen + 1);
return error;
}
/*
* Create a special file.
*/
static int mkfifo1(vfs_context_t ctx, user_addr_t upath, struct vnode_attr *vap,
int fd);
static int
mknodat_internal(proc_t p, user_addr_t upath, struct vnode_attr *vap,
mode_t mode, int fd)
{
vfs_context_t ctx = vfs_context_current();
struct nameidata nd;
vnode_t vp, dvp;
int error;
/* If it's a mknod() of a FIFO, call mkfifo1() instead */
if ((mode & S_IFMT) == S_IFIFO) {
return mkfifo1(ctx, upath, vap, fd);
}
AUDIT_ARG(mode, mode);
AUDIT_ARG(value32, vap->va_rdev);
if ((error = suser(vfs_context_ucred(ctx), &p->p_acflag))) {
return error;
}
NDINIT(&nd, CREATE, OP_MKNOD, LOCKPARENT | AUDITVNPATH1,
UIO_USERSPACE, upath, ctx);
error = nameiat(&nd, fd);
if (error) {
return error;
}
dvp = nd.ni_dvp;
vp = nd.ni_vp;
if (vp != NULL) {
error = EEXIST;
goto out;
}
switch (mode & S_IFMT) {
case S_IFCHR:
VATTR_SET(vap, va_type, VCHR);
break;
case S_IFBLK:
VATTR_SET(vap, va_type, VBLK);
break;
default:
error = EINVAL;
goto out;
}
#if CONFIG_MACF
error = mac_vnode_check_create(ctx,
nd.ni_dvp, &nd.ni_cnd, vap);
if (error) {
goto out;
}
#endif
if ((error = vnode_authorize(dvp, NULL, KAUTH_VNODE_ADD_FILE, ctx)) != 0) {
goto out;
}
#if CONFIG_FILE_LEASES
vnode_breakdirlease(dvp, false, O_WRONLY);
#endif
if ((error = vn_create(dvp, &vp, &nd, vap, 0, 0, NULL, ctx)) != 0) {
goto out;
}
if (vp) {
int update_flags = 0;
// Make sure the name & parent pointers are hooked up
if (vp->v_name == NULL) {
update_flags |= VNODE_UPDATE_NAME;
}
if (vp->v_parent == NULLVP) {
update_flags |= VNODE_UPDATE_PARENT;
}
if (update_flags) {
vnode_update_identity(vp, dvp, nd.ni_cnd.cn_nameptr, nd.ni_cnd.cn_namelen, nd.ni_cnd.cn_hash, update_flags);
}
#if CONFIG_FSE
add_fsevent(FSE_CREATE_FILE, ctx,
FSE_ARG_VNODE, vp,
FSE_ARG_DONE);
#endif
}
out:
/*
* nameidone has to happen before we vnode_put(dvp)
* since it may need to release the fs_nodelock on the dvp
*/
nameidone(&nd);
if (vp) {
vnode_put(vp);
}
vnode_put(dvp);
return error;
}
int
mknod(proc_t p, struct mknod_args *uap, __unused int32_t *retval)
{
struct vnode_attr va;
VATTR_INIT(&va);
VATTR_SET(&va, va_mode, (uap->mode & ALLPERMS) & ~p->p_fd.fd_cmask);
VATTR_SET(&va, va_rdev, uap->dev);
return mknodat_internal(p, uap->path, &va, (mode_t)uap->mode, AT_FDCWD);
}
int
mknodat(proc_t p, struct mknodat_args *uap, __unused int32_t *retval)
{
struct vnode_attr va;
VATTR_INIT(&va);
VATTR_SET(&va, va_mode, (uap->mode & ALLPERMS) & ~p->p_fd.fd_cmask);
VATTR_SET(&va, va_rdev, uap->dev);
return mknodat_internal(p, uap->path, &va, (mode_t)uap->mode, uap->fd);
}
/*
* Create a named pipe.
*
* Returns: 0 Success
* EEXIST
* namei:???
* vnode_authorize:???
* vn_create:???
*/
static int
mkfifo1(vfs_context_t ctx, user_addr_t upath, struct vnode_attr *vap, int fd)
{
vnode_t vp, dvp;
int error;
struct nameidata nd;
NDINIT(&nd, CREATE, OP_MKFIFO, LOCKPARENT | AUDITVNPATH1,
UIO_USERSPACE, upath, ctx);
error = nameiat(&nd, fd);
if (error) {
return error;
}
dvp = nd.ni_dvp;
vp = nd.ni_vp;
/* check that this is a new file and authorize addition */
if (vp != NULL) {
error = EEXIST;
goto out;
}
VATTR_SET(vap, va_type, VFIFO);
if ((error = vn_authorize_create(dvp, &nd.ni_cnd, vap, ctx, NULL)) != 0) {
goto out;
}
error = vn_create(dvp, &vp, &nd, vap, 0, 0, NULL, ctx);
out:
/*
* nameidone has to happen before we vnode_put(dvp)
* since it may need to release the fs_nodelock on the dvp
*/
nameidone(&nd);
if (vp) {
vnode_put(vp);
}
vnode_put(dvp);
return error;
}
/*
* mkfifo_extended: Create a named pipe; with extended argument list (including extended security (ACL)).
*
* Parameters: p Process requesting the open
* uap User argument descriptor (see below)
* retval (Ignored)
*
* Indirect: uap->path Path to fifo (same as 'mkfifo')
* uap->uid UID to set
* uap->gid GID to set
* uap->mode File mode to set (same as 'mkfifo')
* uap->xsecurity ACL to set, if creating
*
* Returns: 0 Success
* !0 errno value
*
* Notes: The kauth_filesec_t in 'va', if any, is in host byte order.
*
* XXX: We should enummerate the possible errno values here, and where
* in the code they originated.
*/
int
mkfifo_extended(proc_t p, struct mkfifo_extended_args *uap, __unused int32_t *retval)
{
int ciferror;
kauth_filesec_t xsecdst;
struct vnode_attr va;
AUDIT_ARG(owner, uap->uid, uap->gid);
xsecdst = KAUTH_FILESEC_NONE;
if (uap->xsecurity != USER_ADDR_NULL) {
if ((ciferror = kauth_copyinfilesec(uap->xsecurity, &xsecdst)) != 0) {
return ciferror;
}
}
VATTR_INIT(&va);
VATTR_SET(&va, va_mode, (uap->mode & ALLPERMS) & ~p->p_fd.fd_cmask);
if (uap->uid != KAUTH_UID_NONE) {
VATTR_SET(&va, va_uid, uap->uid);
}
if (uap->gid != KAUTH_GID_NONE) {
VATTR_SET(&va, va_gid, uap->gid);
}
if (xsecdst != KAUTH_FILESEC_NONE) {
VATTR_SET(&va, va_acl, &xsecdst->fsec_acl);
va.va_vaflags |= VA_FILESEC_ACL;
}
ciferror = mkfifo1(vfs_context_current(), uap->path, &va, AT_FDCWD);
if (xsecdst != KAUTH_FILESEC_NONE) {
kauth_filesec_free(xsecdst);
}
return ciferror;
}
/* ARGSUSED */
int
mkfifo(proc_t p, struct mkfifo_args *uap, __unused int32_t *retval)
{
struct vnode_attr va;
VATTR_INIT(&va);
VATTR_SET(&va, va_mode, (uap->mode & ALLPERMS) & ~p->p_fd.fd_cmask);
return mkfifo1(vfs_context_current(), uap->path, &va, AT_FDCWD);
}
int
mkfifoat(proc_t p, struct mkfifoat_args *uap, __unused int32_t *retval)
{
struct vnode_attr va;
VATTR_INIT(&va);
VATTR_SET(&va, va_mode, (uap->mode & ALLPERMS) & ~p->p_fd.fd_cmask);
return mkfifo1(vfs_context_current(), uap->path, &va, uap->fd);
}
extern int safe_getpath_new(struct vnode *dvp, char *leafname, char *path, int _len, int *truncated_path, int firmlink);
extern int safe_getpath(struct vnode *dvp, char *leafname, char *path, int _len, int *truncated_path);
extern int safe_getpath_no_firmlink(struct vnode *dvp, char *leafname, char *path, int _len, int *truncated_path);
int
safe_getpath_new(struct vnode *dvp, char *leafname, char *path, int _len, int *truncated_path, int firmlink)
{
int ret, len = _len;
*truncated_path = 0;
if (firmlink) {
ret = vn_getpath(dvp, path, &len);
} else {
ret = vn_getpath_no_firmlink(dvp, path, &len);
}
if (ret == 0 && len < (MAXPATHLEN - 1)) {
if (leafname) {
path[len - 1] = '/';
len += strlcpy(&path[len], leafname, MAXPATHLEN - len) + 1;
if (len > MAXPATHLEN) {
char *ptr;
// the string got truncated!
*truncated_path = 1;
ptr = strrchr(path, '/');
if (ptr) {
*ptr = '\0'; // chop off the string at the last directory component
}
len = (int)strlen(path) + 1;
}
}
} else if (ret == 0) {
*truncated_path = 1;
} else if (ret != 0) {
struct vnode *mydvp = dvp;
if (ret != ENOSPC) {
printf("safe_getpath: failed to get the path for vp %p (%s) : err %d\n",
dvp, dvp->v_name ? dvp->v_name : "no-name", ret);
}
*truncated_path = 1;
do {
if (mydvp->v_parent != NULL) {
mydvp = mydvp->v_parent;
} else if (mydvp->v_mount) {
strlcpy(path, mydvp->v_mount->mnt_vfsstat.f_mntonname, _len);
break;
} else {
// no parent and no mount point? only thing is to punt and say "/" changed
strlcpy(path, "/", _len);
len = 2;
mydvp = NULL;
}
if (mydvp == NULL) {
break;
}
len = _len;
if (firmlink) {
ret = vn_getpath(mydvp, path, &len);
} else {
ret = vn_getpath_no_firmlink(mydvp, path, &len);
}
} while (ret == ENOSPC);
}
return len;
}
int
safe_getpath(struct vnode *dvp, char *leafname, char *path, int _len, int *truncated_path)
{
return safe_getpath_new(dvp, leafname, path, _len, truncated_path, 1);
}
int
safe_getpath_no_firmlink(struct vnode *dvp, char *leafname, char *path, int _len, int *truncated_path)
{
return safe_getpath_new(dvp, leafname, path, _len, truncated_path, 0);
}
/*
* Make a hard file link.
*
* Returns: 0 Success
* EPERM
* EEXIST
* EXDEV
* namei:???
* vnode_authorize:???
* VNOP_LINK:???
*/
/* ARGSUSED */
static int
linkat_internal(vfs_context_t ctx, int fd1, user_addr_t path, int fd2,
user_addr_t link, int flag, enum uio_seg segflg)
{
vnode_t vp, pvp, dvp, lvp;
struct nameidata nd;
int follow;
int error;
#if CONFIG_FSE
fse_info finfo;
#endif
int need_event, has_listeners, need_kpath2;
char *target_path = NULL;
char *no_firmlink_path = NULL;
int truncated = 0;
int truncated_no_firmlink_path = 0;
vp = dvp = lvp = NULLVP;
/* look up the object we are linking to */
follow = (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW;
NDINIT(&nd, LOOKUP, OP_LOOKUP, AUDITVNPATH1 | follow,
segflg, path, ctx);
error = nameiat(&nd, fd1);
if (error) {
return error;
}
vp = nd.ni_vp;
nameidone(&nd);
/*
* Normally, linking to directories is not supported.
* However, some file systems may have limited support.
*/
if (vp->v_type == VDIR) {
if (!ISSET(vp->v_mount->mnt_kern_flag, MNTK_DIR_HARDLINKS)) {
error = EPERM; /* POSIX */
goto out;
}
/* Linking to a directory requires ownership. */
if (!kauth_cred_issuser(vfs_context_ucred(ctx))) {
struct vnode_attr dva;
VATTR_INIT(&dva);
VATTR_WANTED(&dva, va_uid);
if (vnode_getattr(vp, &dva, ctx) != 0 ||
!VATTR_IS_SUPPORTED(&dva, va_uid) ||
(dva.va_uid != kauth_cred_getuid(vfs_context_ucred(ctx)))) {
error = EACCES;
goto out;
}
}
}
/* lookup the target node */
#if CONFIG_TRIGGERS
nd.ni_op = OP_LINK;
#endif
nd.ni_cnd.cn_nameiop = CREATE;
nd.ni_cnd.cn_flags = LOCKPARENT | AUDITVNPATH2 | CN_NBMOUNTLOOK;
nd.ni_dirp = link;
error = nameiat(&nd, fd2);
if (error != 0) {
goto out;
}
dvp = nd.ni_dvp;
lvp = nd.ni_vp;
#if CONFIG_MACF
if ((error = mac_vnode_check_link(ctx, dvp, vp, &nd.ni_cnd)) != 0) {
goto out2;
}
#endif
/* or to anything that kauth doesn't want us to (eg. immutable items) */
if ((error = vnode_authorize(vp, NULL, KAUTH_VNODE_LINKTARGET, ctx)) != 0) {
goto out2;
}
/* target node must not exist */
if (lvp != NULLVP) {
error = EEXIST;
goto out2;
}
/* cannot link across mountpoints */
if (vnode_mount(vp) != vnode_mount(dvp)) {
error = EXDEV;
goto out2;
}
/* authorize creation of the target note */
if ((error = vnode_authorize(dvp, NULL, KAUTH_VNODE_ADD_FILE, ctx)) != 0) {
goto out2;
}
#if CONFIG_FILE_LEASES
vnode_breakdirlease(dvp, false, O_WRONLY);
#endif
/* and finally make the link */
error = VNOP_LINK(vp, dvp, &nd.ni_cnd, ctx);
if (error) {
goto out2;
}
#if CONFIG_MACF
(void)mac_vnode_notify_link(ctx, vp, dvp, &nd.ni_cnd);
#endif
#if CONFIG_FSE
need_event = need_fsevent(FSE_CREATE_FILE, dvp);
#else
need_event = 0;
#endif
has_listeners = kauth_authorize_fileop_has_listeners();
need_kpath2 = 0;
#if CONFIG_AUDIT
if (AUDIT_RECORD_EXISTS()) {
need_kpath2 = 1;
}
#endif
if (need_event || has_listeners || need_kpath2) {
char *link_to_path = NULL;
int len, link_name_len;
int len_no_firmlink_path = 0;
/* build the path to the new link file */
GET_PATH(target_path);
len = safe_getpath(dvp, nd.ni_cnd.cn_nameptr, target_path, MAXPATHLEN, &truncated);
if (no_firmlink_path == NULL) {
GET_PATH(no_firmlink_path);
}
len_no_firmlink_path = safe_getpath_no_firmlink(dvp, nd.ni_cnd.cn_nameptr, no_firmlink_path, MAXPATHLEN, &truncated_no_firmlink_path);
AUDIT_ARG(kpath, target_path, ARG_KPATH2);
if (has_listeners) {
/* build the path to file we are linking to */
GET_PATH(link_to_path);
link_name_len = MAXPATHLEN;
if (vn_getpath(vp, link_to_path, &link_name_len) == 0) {
/*
* Call out to allow 3rd party notification of rename.
* Ignore result of kauth_authorize_fileop call.
*/
kauth_authorize_fileop(vfs_context_ucred(ctx), KAUTH_FILEOP_LINK,
(uintptr_t)link_to_path,
(uintptr_t)target_path);
}
if (link_to_path != NULL) {
RELEASE_PATH(link_to_path);
}
}
#if CONFIG_FSE
if (need_event) {
/* construct fsevent */
if (get_fse_info(vp, &finfo, ctx) == 0) {
if (truncated_no_firmlink_path) {
finfo.mode |= FSE_TRUNCATED_PATH;
}
// build the path to the destination of the link
add_fsevent(FSE_CREATE_FILE, ctx,
FSE_ARG_STRING, len_no_firmlink_path, no_firmlink_path,
FSE_ARG_FINFO, &finfo,
FSE_ARG_DONE);
}
pvp = vp->v_parent;
// need an iocount on parent vnode in this case
if (pvp && pvp != dvp) {
pvp = vnode_getparent_if_different(vp, dvp);
}
if (pvp) {
add_fsevent(FSE_STAT_CHANGED, ctx,
FSE_ARG_VNODE, pvp, FSE_ARG_DONE);
}
if (pvp && pvp != dvp) {
vnode_put(pvp);
}
}
#endif
}
out2:
/*
* nameidone has to happen before we vnode_put(dvp)
* since it may need to release the fs_nodelock on the dvp
*/
nameidone(&nd);
if (target_path != NULL) {
RELEASE_PATH(target_path);
}
if (no_firmlink_path != NULL) {
RELEASE_PATH(no_firmlink_path);
no_firmlink_path = NULL;
}
out:
if (lvp) {
vnode_put(lvp);
}
if (dvp) {
vnode_put(dvp);
}
vnode_put(vp);
return error;
}
int
link(__unused proc_t p, struct link_args *uap, __unused int32_t *retval)
{
return linkat_internal(vfs_context_current(), AT_FDCWD, uap->path,
AT_FDCWD, uap->link, AT_SYMLINK_FOLLOW, UIO_USERSPACE);
}
int
linkat(__unused proc_t p, struct linkat_args *uap, __unused int32_t *retval)
{
if (uap->flag & ~AT_SYMLINK_FOLLOW) {
return EINVAL;
}
return linkat_internal(vfs_context_current(), uap->fd1, uap->path,
uap->fd2, uap->link, uap->flag, UIO_USERSPACE);
}
/*
* Make a symbolic link.
*
* We could add support for ACLs here too...
*/
/* ARGSUSED */
static int
symlinkat_internal(vfs_context_t ctx, user_addr_t path_data, int fd,
user_addr_t link, enum uio_seg segflg)
{
struct vnode_attr va;
char *path;
int error;
struct nameidata nd;
vnode_t vp, dvp;
size_t dummy = 0;
proc_t p;
error = 0;
if (UIO_SEG_IS_USER_SPACE(segflg)) {
path = zalloc(ZV_NAMEI);
error = copyinstr(path_data, path, MAXPATHLEN, &dummy);
} else {
path = (char *)path_data;
}
if (error) {
goto out;
}
AUDIT_ARG(text, path); /* This is the link string */
NDINIT(&nd, CREATE, OP_SYMLINK, LOCKPARENT | AUDITVNPATH1,
segflg, link, ctx);
error = nameiat(&nd, fd);
if (error) {
goto out;
}
dvp = nd.ni_dvp;
vp = nd.ni_vp;
p = vfs_context_proc(ctx);
VATTR_INIT(&va);
VATTR_SET(&va, va_type, VLNK);
VATTR_SET(&va, va_mode, ACCESSPERMS & ~p->p_fd.fd_cmask);
#if CONFIG_MACF
error = mac_vnode_check_create(ctx,
dvp, &nd.ni_cnd, &va);
#endif
if (error != 0) {
goto skipit;
}
if (vp != NULL) {
error = EEXIST;
goto skipit;
}
/* authorize */
if (error == 0) {
error = vnode_authorize(dvp, NULL, KAUTH_VNODE_ADD_FILE, ctx);
}
/* get default ownership, etc. */
if (error == 0) {
error = vnode_authattr_new(dvp, &va, 0, ctx);
}
#if CONFIG_FILE_LEASES
vnode_breakdirlease(dvp, false, O_WRONLY);
#endif
if (error == 0) {
error = VNOP_SYMLINK(dvp, &vp, &nd.ni_cnd, &va, path, ctx);
}
/* do fallback attribute handling */
if (error == 0 && vp) {
error = vnode_setattr_fallback(vp, &va, ctx);
}
#if CONFIG_MACF
if (error == 0 && vp) {
error = vnode_label(vnode_mount(vp), dvp, vp, &nd.ni_cnd, VNODE_LABEL_CREATE, ctx);
}
#endif
if (error == 0) {
int update_flags = 0;
/*check if a new vnode was created, else try to get one*/
if (vp == NULL) {
nd.ni_cnd.cn_nameiop = LOOKUP;
#if CONFIG_TRIGGERS
nd.ni_op = OP_LOOKUP;
#endif
/*
* Clear all flags except HASBUF to prevent 'cn_pnbuf' buffer to be
* reallocated again in namei().
*/
nd.ni_cnd.cn_flags &= HASBUF;
error = nameiat(&nd, fd);
if (error) {
goto skipit;
}
vp = nd.ni_vp;
}
#if 0 /* XXX - kauth_todo - is KAUTH_FILEOP_SYMLINK needed? */
/* call out to allow 3rd party notification of rename.
* Ignore result of kauth_authorize_fileop call.
*/
if (kauth_authorize_fileop_has_listeners() &&
namei(&nd) == 0) {
char *new_link_path = NULL;
int len;
/* build the path to the new link file */
new_link_path = get_pathbuff();
len = MAXPATHLEN;
vn_getpath(dvp, new_link_path, &len);
if ((len + 1 + nd.ni_cnd.cn_namelen + 1) < MAXPATHLEN) {
new_link_path[len - 1] = '/';
strlcpy(&new_link_path[len], nd.ni_cnd.cn_nameptr, MAXPATHLEN - len);
}
kauth_authorize_fileop(vfs_context_ucred(ctx), KAUTH_FILEOP_SYMLINK,
(uintptr_t)path, (uintptr_t)new_link_path);
if (new_link_path != NULL) {
release_pathbuff(new_link_path);
}
}
#endif
// Make sure the name & parent pointers are hooked up
if (vp->v_name == NULL) {
update_flags |= VNODE_UPDATE_NAME;
}
if (vp->v_parent == NULLVP) {
update_flags |= VNODE_UPDATE_PARENT;
}
if (update_flags) {
vnode_update_identity(vp, dvp, nd.ni_cnd.cn_nameptr, nd.ni_cnd.cn_namelen, nd.ni_cnd.cn_hash, update_flags);
}
#if CONFIG_FSE
add_fsevent(FSE_CREATE_FILE, ctx,
FSE_ARG_VNODE, vp,
FSE_ARG_DONE);
#endif
}
skipit:
/*
* nameidone has to happen before we vnode_put(dvp)
* since it may need to release the fs_nodelock on the dvp
*/
nameidone(&nd);
if (vp) {
vnode_put(vp);
}
vnode_put(dvp);
out:
if (path && (path != (char *)path_data)) {
zfree(ZV_NAMEI, path);
}
return error;
}
int
symlink(__unused proc_t p, struct symlink_args *uap, __unused int32_t *retval)
{
return symlinkat_internal(vfs_context_current(), uap->path, AT_FDCWD,
uap->link, UIO_USERSPACE);
}
int
symlinkat(__unused proc_t p, struct symlinkat_args *uap,
__unused int32_t *retval)
{
return symlinkat_internal(vfs_context_current(), uap->path1, uap->fd,
uap->path2, UIO_USERSPACE);
}
/*
* Delete a whiteout from the filesystem.
* No longer supported.
*/
int
undelete(__unused proc_t p, __unused struct undelete_args *uap, __unused int32_t *retval)
{
return ENOTSUP;
}
/*
* Delete a name from the filesystem.
*/
/* ARGSUSED */
static int
unlinkat_internal(vfs_context_t ctx, int fd, vnode_t start_dvp,
user_addr_t path_arg, enum uio_seg segflg, int unlink_flags)
{
struct {
struct nameidata nd;
#if CONFIG_FSE
struct vnode_attr va;
fse_info finfo;
#endif
} *__unlink_data;
struct nameidata *ndp;
vnode_t vp, dvp;
int error;
struct componentname *cnp;
char *path = NULL;
char *no_firmlink_path = NULL;
int len_path = 0;
int len_no_firmlink_path = 0;
int flags;
int need_event;
int has_listeners;
int truncated_path;
int truncated_no_firmlink_path;
int batched;
struct vnode_attr *vap;
int do_retry;
int retry_count = 0;
int cn_flags;
int nofollow_any = 0;
cn_flags = LOCKPARENT;
if (!(unlink_flags & VNODE_REMOVE_NO_AUDIT_PATH)) {
cn_flags |= AUDITVNPATH1;
}
if (unlink_flags & VNODE_REMOVE_NOFOLLOW_ANY) {
nofollow_any = NAMEI_NOFOLLOW_ANY;
unlink_flags &= ~VNODE_REMOVE_NOFOLLOW_ANY;
}
/* If a starting dvp is passed, it trumps any fd passed. */
if (start_dvp) {
cn_flags |= USEDVP;
}
#if NAMEDRSRCFORK
/* unlink or delete is allowed on rsrc forks and named streams */
cn_flags |= CN_ALLOWRSRCFORK;
#endif
__unlink_data = kalloc_type(typeof(*__unlink_data), Z_WAITOK);
ndp = &__unlink_data->nd;
#if CONFIG_FSE
fse_info *finfop = &__unlink_data->finfo;
#endif
retry:
do_retry = 0;
flags = 0;
need_event = 0;
has_listeners = 0;
truncated_path = 0;
truncated_no_firmlink_path = 0;
vap = NULL;
NDINIT(ndp, DELETE, OP_UNLINK, cn_flags, segflg, path_arg, ctx);
ndp->ni_dvp = start_dvp;
ndp->ni_flag |= NAMEI_COMPOUNDREMOVE | nofollow_any;
cnp = &ndp->ni_cnd;
continue_lookup:
error = nameiat(ndp, fd);
if (error) {
goto early_out;
}
dvp = ndp->ni_dvp;
vp = ndp->ni_vp;
/* With Carbon delete semantics, busy files cannot be deleted */
if (unlink_flags & VNODE_REMOVE_NODELETEBUSY) {
flags |= VNODE_REMOVE_NODELETEBUSY;
}
/* Skip any potential upcalls if told to. */
if (unlink_flags & VNODE_REMOVE_SKIP_NAMESPACE_EVENT) {
flags |= VNODE_REMOVE_SKIP_NAMESPACE_EVENT;
}
if (vp) {
batched = vnode_compound_remove_available(vp);
/*
* The root of a mounted filesystem cannot be deleted.
*/
if ((vp->v_flag & VROOT) || (dvp->v_mount != vp->v_mount)) {
error = EBUSY;
goto out;
}
#if DEVELOPMENT || DEBUG
/*
* XXX VSWAP: Check for entitlements or special flag here
* so we can restrict access appropriately.
*/
#else /* DEVELOPMENT || DEBUG */
if (vnode_isswap(vp) && (ctx != vfs_context_kernel())) {
error = EPERM;
goto out;
}
#endif /* DEVELOPMENT || DEBUG */
if (!batched) {
error = vn_authorize_unlink(dvp, vp, cnp, ctx, NULL);
if (error) {
if (error == ENOENT) {
if (retry_count < MAX_AUTHORIZE_ENOENT_RETRIES) {
do_retry = 1;
retry_count++;
}
}
goto out;
}
}
} else {
batched = 1;
if (!vnode_compound_remove_available(dvp)) {
panic("No vp, but no compound remove?");
}
}
#if CONFIG_FSE
need_event = need_fsevent(FSE_DELETE, dvp);
if (need_event) {
if (!batched) {
if ((vp->v_flag & VISHARDLINK) == 0) {
/* XXX need to get these data in batched VNOP */
get_fse_info(vp, finfop, ctx);
}
} else {
error =
vfs_get_notify_attributes(&__unlink_data->va);
if (error) {
goto out;
}
vap = &__unlink_data->va;
}
}
#endif
has_listeners = kauth_authorize_fileop_has_listeners();
if (need_event || has_listeners) {
if (path == NULL) {
GET_PATH(path);
}
len_path = safe_getpath(dvp, ndp->ni_cnd.cn_nameptr, path, MAXPATHLEN, &truncated_path);
if (no_firmlink_path == NULL) {
GET_PATH(no_firmlink_path);
}
len_no_firmlink_path = safe_getpath_no_firmlink(dvp, ndp->ni_cnd.cn_nameptr, no_firmlink_path, MAXPATHLEN, &truncated_no_firmlink_path);
}
#if NAMEDRSRCFORK
if (ndp->ni_cnd.cn_flags & CN_WANTSRSRCFORK) {
error = vnode_removenamedstream(dvp, vp, XATTR_RESOURCEFORK_NAME, 0, ctx);
} else
#endif
{
#if CONFIG_FILE_LEASES
vnode_breakdirlease(dvp, false, O_WRONLY);
#endif
error = vn_remove(dvp, &ndp->ni_vp, ndp, flags, vap, ctx);
vp = ndp->ni_vp;
if (error == EKEEPLOOKING) {
if (!batched) {
panic("EKEEPLOOKING, but not a filesystem that supports compound VNOPs?");
}
if ((ndp->ni_flag & NAMEI_CONTLOOKUP) == 0) {
panic("EKEEPLOOKING, but continue flag not set?");
}
if (vnode_isdir(vp)) {
error = EISDIR;
goto out;
}
goto continue_lookup;
} else if (error == ENOENT && batched) {
if (retry_count < MAX_AUTHORIZE_ENOENT_RETRIES) {
/*
* For compound VNOPs, the authorization callback may
* return ENOENT in case of racing hardlink lookups
* hitting the name cache, redrive the lookup.
*/
do_retry = 1;
retry_count += 1;
goto out;
}
}
}
/*
* Call out to allow 3rd party notification of delete.
* Ignore result of kauth_authorize_fileop call.
*/
if (!error) {
if (has_listeners) {
kauth_authorize_fileop(vfs_context_ucred(ctx),
KAUTH_FILEOP_DELETE,
(uintptr_t)vp,
(uintptr_t)path);
}
if (vp->v_flag & VISHARDLINK) {
//
// if a hardlink gets deleted we want to blow away the
// v_parent link because the path that got us to this
// instance of the link is no longer valid. this will
// force the next call to get the path to ask the file
// system instead of just following the v_parent link.
//
vnode_update_identity(vp, NULL, NULL, 0, 0, VNODE_UPDATE_PARENT);
}
#if CONFIG_FSE
if (need_event) {
if (vp->v_flag & VISHARDLINK) {
get_fse_info(vp, finfop, ctx);
} else if (vap) {
vnode_get_fse_info_from_vap(vp, finfop, vap);
}
if (truncated_path) {
finfop->mode |= FSE_TRUNCATED_PATH;
}
add_fsevent(FSE_DELETE, ctx,
FSE_ARG_STRING, len_no_firmlink_path, no_firmlink_path,
FSE_ARG_FINFO, finfop,
FSE_ARG_DONE);
}
#endif
#if CONFIG_MACF
mac_vnode_notify_unlink(ctx, dvp, vp, cnp);
#endif
}
out:
if (path != NULL) {
RELEASE_PATH(path);
path = NULL;
}
if (no_firmlink_path != NULL) {
RELEASE_PATH(no_firmlink_path);
no_firmlink_path = NULL;
}
#if NAMEDRSRCFORK
/* recycle the deleted rsrc fork vnode to force a reclaim, which
* will cause its shadow file to go away if necessary.
*/
if (vp && (vnode_isnamedstream(vp)) &&
(vp->v_parent != NULLVP) &&
vnode_isshadow(vp)) {
vnode_recycle(vp);
}
#endif
/*
* nameidone has to happen before we vnode_put(dvp)
* since it may need to release the fs_nodelock on the dvp
*/
nameidone(ndp);
vnode_put(dvp);
if (vp) {
vnode_put(vp);
}
if (do_retry) {
goto retry;
}
early_out:
kfree_type(typeof(*__unlink_data), __unlink_data);
return error;
}
int
unlink1(vfs_context_t ctx, vnode_t start_dvp, user_addr_t path_arg,
enum uio_seg segflg, int unlink_flags)
{
return unlinkat_internal(ctx, AT_FDCWD, start_dvp, path_arg, segflg,
unlink_flags);
}
/*
* Delete a name from the filesystem using Carbon semantics.
*/
int
delete(__unused proc_t p, struct delete_args *uap, __unused int32_t *retval)
{
return unlinkat_internal(vfs_context_current(), AT_FDCWD, NULLVP,
uap->path, UIO_USERSPACE, VNODE_REMOVE_NODELETEBUSY);
}
/*
* Delete a name from the filesystem using POSIX semantics.
*/
int
unlink(__unused proc_t p, struct unlink_args *uap, __unused int32_t *retval)
{
return unlinkat_internal(vfs_context_current(), AT_FDCWD, NULLVP,
uap->path, UIO_USERSPACE, 0);
}
int
unlinkat(__unused proc_t p, struct unlinkat_args *uap, __unused int32_t *retval)
{
int unlink_flags = 0;
if (uap->flag & ~(AT_REMOVEDIR | AT_REMOVEDIR_DATALESS | AT_SYMLINK_NOFOLLOW_ANY)) {
return EINVAL;
}
if (uap->flag & AT_SYMLINK_NOFOLLOW_ANY) {
unlink_flags |= VNODE_REMOVE_NOFOLLOW_ANY;
}
if (uap->flag & (AT_REMOVEDIR | AT_REMOVEDIR_DATALESS)) {
if (uap->flag & AT_REMOVEDIR_DATALESS) {
unlink_flags |= VNODE_REMOVE_DATALESS_DIR;
}
return rmdirat_internal(vfs_context_current(), uap->fd,
uap->path, UIO_USERSPACE, unlink_flags);
} else {
return unlinkat_internal(vfs_context_current(), uap->fd,
NULLVP, uap->path, UIO_USERSPACE, unlink_flags);
}
}
/*
* Reposition read/write file offset.
*/
int
lseek(proc_t p, struct lseek_args *uap, off_t *retval)
{
struct fileproc *fp;
vnode_t vp;
struct vfs_context *ctx;
off_t offset = uap->offset, file_size;
int error;
if ((error = fp_getfvp(p, uap->fd, &fp, &vp))) {
if (error == ENOTSUP) {
return ESPIPE;
}
return error;
}
if (vnode_isfifo(vp)) {
file_drop(uap->fd);
return ESPIPE;
}
ctx = vfs_context_current();
#if CONFIG_MACF
if (uap->whence == L_INCR && uap->offset == 0) {
error = mac_file_check_get_offset(vfs_context_ucred(ctx),
fp->fp_glob);
} else {
error = mac_file_check_change_offset(vfs_context_ucred(ctx),
fp->fp_glob);
}
if (error) {
file_drop(uap->fd);
return error;
}
#endif
if ((error = vnode_getwithref(vp))) {
file_drop(uap->fd);
return error;
}
switch (uap->whence) {
case L_INCR:
offset += fp->fp_glob->fg_offset;
break;
case L_XTND:
if ((error = vnode_size(vp, &file_size, ctx)) != 0) {
break;
}
offset += file_size;
break;
case L_SET:
break;
case SEEK_HOLE:
error = VNOP_IOCTL(vp, FSIOC_FIOSEEKHOLE, (caddr_t)&offset, 0, ctx);
break;
case SEEK_DATA:
error = VNOP_IOCTL(vp, FSIOC_FIOSEEKDATA, (caddr_t)&offset, 0, ctx);
break;
default:
error = EINVAL;
}
if (error == 0) {
if (uap->offset > 0 && offset < 0) {
/* Incremented/relative move past max size */
error = EOVERFLOW;
} else {
/*
* Allow negative offsets on character devices, per
* POSIX 1003.1-2001. Most likely for writing disk
* labels.
*/
if (offset < 0 && vp->v_type != VCHR) {
/* Decremented/relative move before start */
error = EINVAL;
} else {
/* Success */
fp->fp_glob->fg_offset = offset;
*retval = fp->fp_glob->fg_offset;
}
}
}
/*
* An lseek can affect whether data is "available to read." Use
* hint of NOTE_NONE so no EVFILT_VNODE events fire
*/
post_event_if_success(vp, error, NOTE_NONE);
(void)vnode_put(vp);
file_drop(uap->fd);
return error;
}
/*
* Check access permissions.
*
* Returns: 0 Success
* vnode_authorize:???
*/
static int
access1(vnode_t vp, vnode_t dvp, int uflags, vfs_context_t ctx)
{
kauth_action_t action;
int error;
/*
* If just the regular access bits, convert them to something
* that vnode_authorize will understand.
*/
if (!(uflags & _ACCESS_EXTENDED_MASK)) {
action = 0;
if (uflags & R_OK) {
action |= KAUTH_VNODE_READ_DATA; /* aka KAUTH_VNODE_LIST_DIRECTORY */
}
if (uflags & W_OK) {
if (vnode_isdir(vp)) {
action |= KAUTH_VNODE_ADD_FILE |
KAUTH_VNODE_ADD_SUBDIRECTORY;
/* might want delete rights here too */
} else {
action |= KAUTH_VNODE_WRITE_DATA;
}
}
if (uflags & X_OK) {
if (vnode_isdir(vp)) {
action |= KAUTH_VNODE_SEARCH;
} else {
action |= KAUTH_VNODE_EXECUTE;
}
}
} else {
/* take advantage of definition of uflags */
action = uflags >> 8;
}
#if CONFIG_MACF
error = mac_vnode_check_access(ctx, vp, uflags);
if (error) {
return error;
}
#endif /* MAC */
/* action == 0 means only check for existence */
if (action != 0) {
error = vnode_authorize(vp, dvp, action | KAUTH_VNODE_ACCESS, ctx);
} else {
error = 0;
}
return error;
}
/*
* access_extended: Check access permissions in bulk.
*
* Description: uap->entries Pointer to an array of accessx
* descriptor structs, plus one or
* more NULL terminated strings (see
* "Notes" section below).
* uap->size Size of the area pointed to by
* uap->entries.
* uap->results Pointer to the results array.
*
* Returns: 0 Success
* ENOMEM Insufficient memory
* EINVAL Invalid arguments
* namei:EFAULT Bad address
* namei:ENAMETOOLONG Filename too long
* namei:ENOENT No such file or directory
* namei:ELOOP Too many levels of symbolic links
* namei:EBADF Bad file descriptor
* namei:ENOTDIR Not a directory
* namei:???
* access1:
*
* Implicit returns:
* uap->results Array contents modified
*
* Notes: The uap->entries are structured as an arbitrary length array
* of accessx descriptors, followed by one or more NULL terminated
* strings
*
* struct accessx_descriptor[0]
* ...
* struct accessx_descriptor[n]
* char name_data[0];
*
* We determine the entry count by walking the buffer containing
* the uap->entries argument descriptor. For each descriptor we
* see, the valid values for the offset ad_name_offset will be
* in the byte range:
*
* [ uap->entries + sizeof(struct accessx_descriptor) ]
* to
* [ uap->entries + uap->size - 2 ]
*
* since we must have at least one string, and the string must
* be at least one character plus the NULL terminator in length.
*
* XXX: Need to support the check-as uid argument
*/
int
access_extended(__unused proc_t p, struct access_extended_args *uap, __unused int32_t *retval)
{
struct accessx_descriptor *input = NULL;
errno_t *result = NULL;
errno_t error = 0;
int wantdelete = 0;
size_t desc_max, desc_actual = 0;
unsigned int i, j;
struct vfs_context context;
struct nameidata nd;
int niopts;
vnode_t vp = NULL;
vnode_t dvp = NULL;
#define ACCESSX_MAX_DESCR_ON_STACK 10
struct accessx_descriptor stack_input[ACCESSX_MAX_DESCR_ON_STACK];
context.vc_ucred = NULL;
/*
* Validate parameters; if valid, copy the descriptor array and string
* arguments into local memory. Before proceeding, the following
* conditions must have been met:
*
* o The total size is not permitted to exceed ACCESSX_MAX_TABLESIZE
* o There must be sufficient room in the request for at least one
* descriptor and a one yte NUL terminated string.
* o The allocation of local storage must not fail.
*/
if (uap->size > ACCESSX_MAX_TABLESIZE) {
return ENOMEM;
}
if (uap->size < (sizeof(struct accessx_descriptor) + 2)) {
return EINVAL;
}
if (uap->size <= sizeof(stack_input)) {
input = stack_input;
} else {
input = kalloc_data(uap->size, Z_WAITOK);
if (input == NULL) {
error = ENOMEM;
goto out;
}
}
error = copyin(uap->entries, input, uap->size);
if (error) {
goto out;
}
AUDIT_ARG(opaque, input, uap->size);
/*
* Force NUL termination of the copyin buffer to avoid nami() running
* off the end. If the caller passes us bogus data, they may get a
* bogus result.
*/
((char *)input)[uap->size - 1] = 0;
/*
* Access is defined as checking against the process' real identity,
* even if operations are checking the effective identity. This
* requires that we use a local vfs context.
*/
context.vc_ucred = kauth_cred_copy_real(kauth_cred_get());
context.vc_thread = current_thread();
/*
* Find out how many entries we have, so we can allocate the result
* array by walking the list and adjusting the count downward by the
* earliest string offset we see.
*/
desc_max = (uap->size - 2) / sizeof(struct accessx_descriptor);
desc_actual = desc_max;
for (i = 0; i < desc_actual; i++) {
/*
* Take the offset to the name string for this entry and
* convert to an input array index, which would be one off
* the end of the array if this entry was the lowest-addressed
* name string.
*/
j = input[i].ad_name_offset / sizeof(struct accessx_descriptor);
/*
* An offset greater than the max allowable offset is an error.
* It is also an error for any valid entry to point
* to a location prior to the end of the current entry, if
* it's not a reference to the string of the previous entry.
*/
if (j > desc_max || (j != 0 && j <= i)) {
error = EINVAL;
goto out;
}
/* Also do not let ad_name_offset point to something beyond the size of the input */
if (input[i].ad_name_offset >= uap->size) {
error = EINVAL;
goto out;
}
/*
* An offset of 0 means use the previous descriptor's offset;
* this is used to chain multiple requests for the same file
* to avoid multiple lookups.
*/
if (j == 0) {
/* This is not valid for the first entry */
if (i == 0) {
error = EINVAL;
goto out;
}
continue;
}
/*
* If the offset of the string for this descriptor is before
* what we believe is the current actual last descriptor,
* then we need to adjust our estimate downward; this permits
* the string table following the last descriptor to be out
* of order relative to the descriptor list.
*/
if (j < desc_actual) {
desc_actual = j;
}
}
/*
* We limit the actual number of descriptors we are willing to process
* to a hard maximum of ACCESSX_MAX_DESCRIPTORS. If the number being
* requested does not exceed this limit,
*/
if (desc_actual > ACCESSX_MAX_DESCRIPTORS) {
error = ENOMEM;
goto out;
}
result = kalloc_data(desc_actual * sizeof(errno_t), Z_WAITOK | Z_ZERO);
if (result == NULL) {
error = ENOMEM;
goto out;
}
/*
* Do the work by iterating over the descriptor entries we know to
* at least appear to contain valid data.
*/
error = 0;
for (i = 0; i < desc_actual; i++) {
/*
* If the ad_name_offset is 0, then we use the previous
* results to make the check; otherwise, we are looking up
* a new file name.
*/
if (input[i].ad_name_offset != 0) {
/* discard old vnodes */
if (vp) {
vnode_put(vp);
vp = NULL;
}
if (dvp) {
vnode_put(dvp);
dvp = NULL;
}
/*
* Scan forward in the descriptor list to see if we
* need the parent vnode. We will need it if we are
* deleting, since we must have rights to remove
* entries in the parent directory, as well as the
* rights to delete the object itself.
*/
wantdelete = input[i].ad_flags & _DELETE_OK;
for (j = i + 1; (j < desc_actual) && (input[j].ad_name_offset == 0); j++) {
if (input[j].ad_flags & _DELETE_OK) {
wantdelete = 1;
}
}
niopts = FOLLOW | AUDITVNPATH1;
/* need parent for vnode_authorize for deletion test */
if (wantdelete) {
niopts |= WANTPARENT;
}
/* do the lookup */
NDINIT(&nd, LOOKUP, OP_ACCESS, niopts, UIO_SYSSPACE,
CAST_USER_ADDR_T(((const char *)input) + input[i].ad_name_offset),
&context);
error = namei(&nd);
if (!error) {
vp = nd.ni_vp;
if (wantdelete) {
dvp = nd.ni_dvp;
}
}
nameidone(&nd);
}
/*
* Handle lookup errors.
*/
switch (error) {
case ENOENT:
case EACCES:
case EPERM:
case ENOTDIR:
result[i] = error;
break;
case 0:
/* run this access check */
result[i] = access1(vp, dvp, input[i].ad_flags, &context);
break;
default:
/* fatal lookup error */
goto out;
}
}
AUDIT_ARG(data, result, sizeof(errno_t), desc_actual);
/* copy out results */
error = copyout(result, uap->results, desc_actual * sizeof(errno_t));
out:
if (input && input != stack_input) {
kfree_data(input, uap->size);
}
if (result) {
kfree_data(result, desc_actual * sizeof(errno_t));
}
if (vp) {
vnode_put(vp);
}
if (dvp) {
vnode_put(dvp);
}
if (IS_VALID_CRED(context.vc_ucred)) {
kauth_cred_unref(&context.vc_ucred);
}
return error;
}
/*
* Returns: 0 Success
* namei:EFAULT Bad address
* namei:ENAMETOOLONG Filename too long
* namei:ENOENT No such file or directory
* namei:ELOOP Too many levels of symbolic links
* namei:EBADF Bad file descriptor
* namei:ENOTDIR Not a directory
* namei:???
* access1:
*/
static int
faccessat_internal(vfs_context_t ctx, int fd, user_addr_t path, int amode,
int flag, enum uio_seg segflg)
{
int error;
struct nameidata nd;
int niopts;
struct vfs_context context;
#if NAMEDRSRCFORK
int is_namedstream = 0;
#endif
/*
* Unless the AT_EACCESS option is used, Access is defined as checking
* against the process' real identity, even if operations are checking
* the effective identity. So we need to tweak the credential
* in the context for that case.
*/
if (!(flag & AT_EACCESS)) {
context.vc_ucred = kauth_cred_copy_real(kauth_cred_get());
} else {
context.vc_ucred = ctx->vc_ucred;
}
context.vc_thread = ctx->vc_thread;
niopts = (flag & (AT_SYMLINK_NOFOLLOW | AT_SYMLINK_NOFOLLOW_ANY) ? NOFOLLOW : FOLLOW) | AUDITVNPATH1;
/* need parent for vnode_authorize for deletion test */
if (amode & _DELETE_OK) {
niopts |= WANTPARENT;
}
NDINIT(&nd, LOOKUP, OP_ACCESS, niopts, segflg,
path, &context);
if (flag & AT_SYMLINK_NOFOLLOW_ANY) {
nd.ni_flag |= NAMEI_NOFOLLOW_ANY;
}
#if NAMEDRSRCFORK
/* access(F_OK) calls are allowed for resource forks. */
if (amode == F_OK) {
nd.ni_cnd.cn_flags |= CN_ALLOWRSRCFORK;
}
#endif
error = nameiat(&nd, fd);
if (error) {
goto out;
}
#if NAMEDRSRCFORK
/* Grab reference on the shadow stream file vnode to
* force an inactive on release which will mark it
* for recycle.
*/
if (vnode_isnamedstream(nd.ni_vp) &&
(nd.ni_vp->v_parent != NULLVP) &&
vnode_isshadow(nd.ni_vp)) {
is_namedstream = 1;
vnode_ref(nd.ni_vp);
}
#endif
error = access1(nd.ni_vp, nd.ni_dvp, amode, &context);
#if NAMEDRSRCFORK
if (is_namedstream) {
vnode_rele(nd.ni_vp);
}
#endif
vnode_put(nd.ni_vp);
if (amode & _DELETE_OK) {
vnode_put(nd.ni_dvp);
}
nameidone(&nd);
out:
if (!(flag & AT_EACCESS)) {
kauth_cred_unref(&context.vc_ucred);
}
return error;
}
int
access(__unused proc_t p, struct access_args *uap, __unused int32_t *retval)
{
return faccessat_internal(vfs_context_current(), AT_FDCWD,
uap->path, uap->flags, 0, UIO_USERSPACE);
}
int
faccessat(__unused proc_t p, struct faccessat_args *uap,
__unused int32_t *retval)
{
if (uap->flag & ~(AT_EACCESS | AT_SYMLINK_NOFOLLOW | AT_SYMLINK_NOFOLLOW_ANY)) {
return EINVAL;
}
return faccessat_internal(vfs_context_current(), uap->fd,
uap->path, uap->amode, uap->flag, UIO_USERSPACE);
}
/*
* Returns: 0 Success
* EFAULT
* copyout:EFAULT
* namei:???
* vn_stat:???
*/
static int
fstatat_internal(vfs_context_t ctx, user_addr_t path, user_addr_t ub,
user_addr_t xsecurity, user_addr_t xsecurity_size, int isstat64,
enum uio_seg segflg, int fd, int flag)
{
struct nameidata *ndp = NULL;
int follow;
union {
struct stat sb;
struct stat64 sb64;
} source = {};
union {
struct user64_stat user64_sb;
struct user32_stat user32_sb;
struct user64_stat64 user64_sb64;
struct user32_stat64 user32_sb64;
} dest = {};
caddr_t sbp;
int error, my_size;
kauth_filesec_t fsec = KAUTH_FILESEC_NONE;
size_t xsecurity_bufsize;
void * statptr;
struct fileproc *fp = NULL;
int needsrealdev = 0;
follow = (flag & (AT_SYMLINK_NOFOLLOW | AT_SYMLINK_NOFOLLOW_ANY)) ? NOFOLLOW : FOLLOW;
ndp = kalloc_type(struct nameidata, Z_WAITOK);
NDINIT(ndp, LOOKUP, OP_GETATTR, follow | AUDITVNPATH1,
segflg, path, ctx);
if (flag & AT_SYMLINK_NOFOLLOW_ANY) {
ndp->ni_flag |= NAMEI_NOFOLLOW_ANY;
}
#if NAMEDRSRCFORK
int is_namedstream = 0;
/* stat calls are allowed for resource forks. */
ndp->ni_cnd.cn_flags |= CN_ALLOWRSRCFORK;
#endif
if (flag & AT_FDONLY) {
vnode_t fvp;
error = fp_getfvp(vfs_context_proc(ctx), fd, &fp, &fvp);
if (error) {
goto out;
}
if ((error = vnode_getwithref(fvp))) {
file_drop(fd);
goto out;
}
ndp->ni_vp = fvp;
} else {
error = nameiat(ndp, fd);
if (error) {
goto out;
}
}
statptr = (void *)&source;
#if NAMEDRSRCFORK
/* Grab reference on the shadow stream file vnode to
* force an inactive on release which will mark it
* for recycle.
*/
if (vnode_isnamedstream(ndp->ni_vp) &&
(ndp->ni_vp->v_parent != NULLVP) &&
vnode_isshadow(ndp->ni_vp)) {
is_namedstream = 1;
vnode_ref(ndp->ni_vp);
}
#endif
needsrealdev = flag & AT_REALDEV ? 1 : 0;
if (fp && (xsecurity == USER_ADDR_NULL)) {
/*
* If the caller has the file open, and is not
* requesting extended security information, we are
* going to let them get the basic stat information.
*/
error = vn_stat_noauth(ndp->ni_vp, statptr, NULL, isstat64, needsrealdev, ctx,
fp->fp_glob->fg_cred);
} else {
error = vn_stat(ndp->ni_vp, statptr, (xsecurity != USER_ADDR_NULL ? &fsec : NULL),
isstat64, needsrealdev, ctx);
}
#if NAMEDRSRCFORK
if (is_namedstream) {
vnode_rele(ndp->ni_vp);
}
#endif
vnode_put(ndp->ni_vp);
nameidone(ndp);
if (fp) {
file_drop(fd);
fp = NULL;
}
if (error) {
goto out;
}
/* Zap spare fields */
if (isstat64 != 0) {
source.sb64.st_lspare = 0;
source.sb64.st_qspare[0] = 0LL;
source.sb64.st_qspare[1] = 0LL;
if (vfs_context_is64bit(ctx)) {
munge_user64_stat64(&source.sb64, &dest.user64_sb64);
my_size = sizeof(dest.user64_sb64);
sbp = (caddr_t)&dest.user64_sb64;
} else {
munge_user32_stat64(&source.sb64, &dest.user32_sb64);
my_size = sizeof(dest.user32_sb64);
sbp = (caddr_t)&dest.user32_sb64;
}
/*
* Check if we raced (post lookup) against the last unlink of a file.
*/
if ((source.sb64.st_nlink == 0) && S_ISREG(source.sb64.st_mode)) {
source.sb64.st_nlink = 1;
}
} else {
source.sb.st_lspare = 0;
source.sb.st_qspare[0] = 0LL;
source.sb.st_qspare[1] = 0LL;
if (vfs_context_is64bit(ctx)) {
munge_user64_stat(&source.sb, &dest.user64_sb);
my_size = sizeof(dest.user64_sb);
sbp = (caddr_t)&dest.user64_sb;
} else {
munge_user32_stat(&source.sb, &dest.user32_sb);
my_size = sizeof(dest.user32_sb);
sbp = (caddr_t)&dest.user32_sb;
}
/*
* Check if we raced (post lookup) against the last unlink of a file.
*/
if ((source.sb.st_nlink == 0) && S_ISREG(source.sb.st_mode)) {
source.sb.st_nlink = 1;
}
}
if ((error = copyout(sbp, ub, my_size)) != 0) {
goto out;
}
/* caller wants extended security information? */
if (xsecurity != USER_ADDR_NULL) {
/* did we get any? */
if (fsec == KAUTH_FILESEC_NONE) {
if (susize(xsecurity_size, 0) != 0) {
error = EFAULT;
goto out;
}
} else {
/* find the user buffer size */
xsecurity_bufsize = fusize(xsecurity_size);
/* copy out the actual data size */
if (susize(xsecurity_size, KAUTH_FILESEC_COPYSIZE(fsec)) != 0) {
error = EFAULT;
goto out;
}
/* if the caller supplied enough room, copy out to it */
if (xsecurity_bufsize >= KAUTH_FILESEC_COPYSIZE(fsec)) {
error = copyout(fsec, xsecurity, KAUTH_FILESEC_COPYSIZE(fsec));
}
}
}
out:
if (ndp) {
kfree_type(struct nameidata, ndp);
}
if (fsec != KAUTH_FILESEC_NONE) {
kauth_filesec_free(fsec);
}
return error;
}
/*
* stat_extended: Get file status; with extended security (ACL).
*
* Parameters: p (ignored)
* uap User argument descriptor (see below)
* retval (ignored)
*
* Indirect: uap->path Path of file to get status from
* uap->ub User buffer (holds file status info)
* uap->xsecurity ACL to get (extended security)
* uap->xsecurity_size Size of ACL
*
* Returns: 0 Success
* !0 errno value
*
*/
int
stat_extended(__unused proc_t p, struct stat_extended_args *uap,
__unused int32_t *retval)
{
return fstatat_internal(vfs_context_current(), uap->path, uap->ub,
uap->xsecurity, uap->xsecurity_size, 0, UIO_USERSPACE, AT_FDCWD,
0);
}
/*
* Returns: 0 Success
* fstatat_internal:??? [see fstatat_internal() in this file]
*/
int
stat(__unused proc_t p, struct stat_args *uap, __unused int32_t *retval)
{
return fstatat_internal(vfs_context_current(), uap->path, uap->ub,
0, 0, 0, UIO_USERSPACE, AT_FDCWD, 0);
}
int
stat64(__unused proc_t p, struct stat64_args *uap, __unused int32_t *retval)
{
return fstatat_internal(vfs_context_current(), uap->path, uap->ub,
0, 0, 1, UIO_USERSPACE, AT_FDCWD, 0);
}
/*
* stat64_extended: Get file status; can handle large inode numbers; with extended security (ACL).
*
* Parameters: p (ignored)
* uap User argument descriptor (see below)
* retval (ignored)
*
* Indirect: uap->path Path of file to get status from
* uap->ub User buffer (holds file status info)
* uap->xsecurity ACL to get (extended security)
* uap->xsecurity_size Size of ACL
*
* Returns: 0 Success
* !0 errno value
*
*/
int
stat64_extended(__unused proc_t p, struct stat64_extended_args *uap, __unused int32_t *retval)
{
return fstatat_internal(vfs_context_current(), uap->path, uap->ub,
uap->xsecurity, uap->xsecurity_size, 1, UIO_USERSPACE, AT_FDCWD,
0);
}
/*
* lstat_extended: Get file status; does not follow links; with extended security (ACL).
*
* Parameters: p (ignored)
* uap User argument descriptor (see below)
* retval (ignored)
*
* Indirect: uap->path Path of file to get status from
* uap->ub User buffer (holds file status info)
* uap->xsecurity ACL to get (extended security)
* uap->xsecurity_size Size of ACL
*
* Returns: 0 Success
* !0 errno value
*
*/
int
lstat_extended(__unused proc_t p, struct lstat_extended_args *uap, __unused int32_t *retval)
{
return fstatat_internal(vfs_context_current(), uap->path, uap->ub,
uap->xsecurity, uap->xsecurity_size, 0, UIO_USERSPACE, AT_FDCWD,
AT_SYMLINK_NOFOLLOW);
}
/*
* Get file status; this version does not follow links.
*/
int
lstat(__unused proc_t p, struct lstat_args *uap, __unused int32_t *retval)
{
return fstatat_internal(vfs_context_current(), uap->path, uap->ub,
0, 0, 0, UIO_USERSPACE, AT_FDCWD, AT_SYMLINK_NOFOLLOW);
}
int
lstat64(__unused proc_t p, struct lstat64_args *uap, __unused int32_t *retval)
{
return fstatat_internal(vfs_context_current(), uap->path, uap->ub,
0, 0, 1, UIO_USERSPACE, AT_FDCWD, AT_SYMLINK_NOFOLLOW);
}
/*
* lstat64_extended: Get file status; can handle large inode numbers; does not
* follow links; with extended security (ACL).
*
* Parameters: p (ignored)
* uap User argument descriptor (see below)
* retval (ignored)
*
* Indirect: uap->path Path of file to get status from
* uap->ub User buffer (holds file status info)
* uap->xsecurity ACL to get (extended security)
* uap->xsecurity_size Size of ACL
*
* Returns: 0 Success
* !0 errno value
*
*/
int
lstat64_extended(__unused proc_t p, struct lstat64_extended_args *uap, __unused int32_t *retval)
{
return fstatat_internal(vfs_context_current(), uap->path, uap->ub,
uap->xsecurity, uap->xsecurity_size, 1, UIO_USERSPACE, AT_FDCWD,
AT_SYMLINK_NOFOLLOW);
}
int
fstatat(__unused proc_t p, struct fstatat_args *uap, __unused int32_t *retval)
{
if (uap->flag & ~(AT_SYMLINK_NOFOLLOW | AT_REALDEV | AT_FDONLY | AT_SYMLINK_NOFOLLOW_ANY)) {
return EINVAL;
}
return fstatat_internal(vfs_context_current(), uap->path, uap->ub,
0, 0, 0, UIO_USERSPACE, uap->fd, uap->flag);
}
int
fstatat64(__unused proc_t p, struct fstatat64_args *uap,
__unused int32_t *retval)
{
if (uap->flag & ~(AT_SYMLINK_NOFOLLOW | AT_REALDEV | AT_FDONLY | AT_SYMLINK_NOFOLLOW_ANY)) {
return EINVAL;
}
return fstatat_internal(vfs_context_current(), uap->path, uap->ub,
0, 0, 1, UIO_USERSPACE, uap->fd, uap->flag);
}
/*
* Get configurable pathname variables.
*
* Returns: 0 Success
* namei:???
* vn_pathconf:???
*
* Notes: Global implementation constants are intended to be
* implemented in this function directly; all other constants
* are per-FS implementation, and therefore must be handled in
* each respective FS, instead.
*
* XXX We implement some things globally right now that should actually be
* XXX per-FS; we will need to deal with this at some point.
*/
/* ARGSUSED */
int
pathconf(__unused proc_t p, struct pathconf_args *uap, int32_t *retval)
{
int error;
struct nameidata nd;
vfs_context_t ctx = vfs_context_current();
NDINIT(&nd, LOOKUP, OP_PATHCONF, FOLLOW | AUDITVNPATH1,
UIO_USERSPACE, uap->path, ctx);
error = namei(&nd);
if (error) {
return error;
}
error = vn_pathconf(nd.ni_vp, uap->name, retval, ctx);
vnode_put(nd.ni_vp);
nameidone(&nd);
return error;
}
/*
* Return target name of a symbolic link.
*/
/* ARGSUSED */
static int
readlinkat_internal(vfs_context_t ctx, int fd, vnode_t lnk_vp, user_addr_t path,
enum uio_seg seg, user_addr_t buf, size_t bufsize, enum uio_seg bufseg,
int *retval)
{
vnode_t vp;
uio_t auio;
int error;
struct nameidata nd;
UIO_STACKBUF(uio_buf, 1);
bool put_vnode;
if (bufsize > INT32_MAX) {
return EINVAL;
}
if (lnk_vp) {
vp = lnk_vp;
put_vnode = false;
} else {
NDINIT(&nd, LOOKUP, OP_READLINK, NOFOLLOW | AUDITVNPATH1,
seg, path, ctx);
error = nameiat(&nd, fd);
if (error) {
return error;
}
vp = nd.ni_vp;
put_vnode = true;
nameidone(&nd);
}
auio = uio_createwithbuffer(1, 0, bufseg, UIO_READ,
&uio_buf[0], sizeof(uio_buf));
uio_addiov(auio, buf, bufsize);
if (vp->v_type != VLNK) {
error = EINVAL;
} else {
#if CONFIG_MACF
error = mac_vnode_check_readlink(ctx, vp);
#endif
if (error == 0) {
error = vnode_authorize(vp, NULL, KAUTH_VNODE_READ_DATA,
ctx);
}
if (error == 0) {
error = VNOP_READLINK(vp, auio, ctx);
}
}
if (put_vnode) {
vnode_put(vp);
}
*retval = (int)(bufsize - uio_resid(auio));
return error;
}
int
freadlink(proc_t p, struct freadlink_args *uap, int32_t *retval)
{
enum uio_seg procseg;
vnode_t vp;
int error;
procseg = IS_64BIT_PROCESS(p) ? UIO_USERSPACE64 : UIO_USERSPACE32;
AUDIT_ARG(fd, uap->fd);
if ((error = file_vnode(uap->fd, &vp))) {
return error;
}
if ((error = vnode_getwithref(vp))) {
file_drop(uap->fd);
return error;
}
error = readlinkat_internal(vfs_context_current(), -1,
vp, 0, procseg, CAST_USER_ADDR_T(uap->buf),
uap->bufsize, procseg, retval);
vnode_put(vp);
file_drop(uap->fd);
return error;
}
int
readlink(proc_t p, struct readlink_args *uap, int32_t *retval)
{
enum uio_seg procseg;
procseg = IS_64BIT_PROCESS(p) ? UIO_USERSPACE64 : UIO_USERSPACE32;
return readlinkat_internal(vfs_context_current(), AT_FDCWD, NULL,
CAST_USER_ADDR_T(uap->path), procseg, CAST_USER_ADDR_T(uap->buf),
uap->count, procseg, retval);
}
int
readlinkat(proc_t p, struct readlinkat_args *uap, int32_t *retval)
{
enum uio_seg procseg;
procseg = IS_64BIT_PROCESS(p) ? UIO_USERSPACE64 : UIO_USERSPACE32;
return readlinkat_internal(vfs_context_current(), uap->fd, NULL,
CAST_USER_ADDR_T(uap->path), procseg, uap->buf, uap->bufsize, procseg,
retval);
}
/*
* Change file flags, the deep inner layer.
*/
static int
chflags0(vnode_t vp, struct vnode_attr *va,
int (*setattr)(vnode_t, void *, vfs_context_t),
void *arg, vfs_context_t ctx)
{
kauth_action_t action = 0;
int error;
#if CONFIG_MACF
error = mac_vnode_check_setflags(ctx, vp, va->va_flags);
if (error) {
goto out;
}
#endif
/* request authorisation, disregard immutability */
if ((error = vnode_authattr(vp, va, &action, ctx)) != 0) {
goto out;
}
/*
* Request that the auth layer disregard those file flags it's allowed to when
* authorizing this operation; we need to do this in order to be able to
* clear immutable flags.
*/
if (action && ((error = vnode_authorize(vp, NULL, action | KAUTH_VNODE_NOIMMUTABLE, ctx)) != 0)) {
goto out;
}
error = (*setattr)(vp, arg, ctx);
#if CONFIG_MACF
if (error == 0) {
mac_vnode_notify_setflags(ctx, vp, va->va_flags);
}
#endif
out:
return error;
}
/*
* Change file flags.
*
* NOTE: this will vnode_put() `vp'
*/
static int
chflags1(vnode_t vp, int flags, vfs_context_t ctx)
{
struct vnode_attr va;
int error;
VATTR_INIT(&va);
VATTR_SET(&va, va_flags, flags);
error = chflags0(vp, &va, (void *)vnode_setattr, &va, ctx);
vnode_put(vp);
if ((error == 0) && !VATTR_IS_SUPPORTED(&va, va_flags)) {
error = ENOTSUP;
}
return error;
}
/*
* Change flags of a file given a path name.
*/
/* ARGSUSED */
int
chflags(__unused proc_t p, struct chflags_args *uap, __unused int32_t *retval)
{
vnode_t vp;
vfs_context_t ctx = vfs_context_current();
int error;
struct nameidata nd;
uint32_t wantparent = 0;
#if CONFIG_FILE_LEASES
wantparent = WANTPARENT;
#endif
AUDIT_ARG(fflags, uap->flags);
NDINIT(&nd, LOOKUP, OP_SETATTR, FOLLOW | AUDITVNPATH1 | wantparent,
UIO_USERSPACE, uap->path, ctx);
error = namei(&nd);
if (error) {
return error;
}
vp = nd.ni_vp;
#if CONFIG_FILE_LEASES
vnode_breakdirlease(nd.ni_dvp, false, O_WRONLY);
vnode_put(nd.ni_dvp);
#endif
nameidone(&nd);
/* we don't vnode_put() here because chflags1 does internally */
error = chflags1(vp, uap->flags, ctx);
return error;
}
/*
* Change flags of a file given a file descriptor.
*/
/* ARGSUSED */
int
fchflags(__unused proc_t p, struct fchflags_args *uap, __unused int32_t *retval)
{
vnode_t vp;
int error;
AUDIT_ARG(fd, uap->fd);
AUDIT_ARG(fflags, uap->flags);
if ((error = file_vnode(uap->fd, &vp))) {
return error;
}
if ((error = vnode_getwithref(vp))) {
file_drop(uap->fd);
return error;
}
AUDIT_ARG(vnpath, vp, ARG_VNODE1);
#if CONFIG_FILE_LEASES
vnode_breakdirlease(vp, true, O_WRONLY);
#endif
/* we don't vnode_put() here because chflags1 does internally */
error = chflags1(vp, uap->flags, vfs_context_current());
file_drop(uap->fd);
return error;
}
/*
* Change security information on a filesystem object.
*
* Returns: 0 Success
* EPERM Operation not permitted
* vnode_authattr:??? [anything vnode_authattr can return]
* vnode_authorize:??? [anything vnode_authorize can return]
* vnode_setattr:??? [anything vnode_setattr can return]
*
* Notes: If vnode_authattr or vnode_authorize return EACCES, it will be
* translated to EPERM before being returned.
*/
static int
chmod_vnode(vfs_context_t ctx, vnode_t vp, struct vnode_attr *vap)
{
kauth_action_t action;
int error;
AUDIT_ARG(mode, vap->va_mode);
/* XXX audit new args */
#if NAMEDSTREAMS
/* chmod calls are not allowed for resource forks. */
if (vp->v_flag & VISNAMEDSTREAM) {
return EPERM;
}
#endif
#if CONFIG_MACF
if (VATTR_IS_ACTIVE(vap, va_mode) &&
(error = mac_vnode_check_setmode(ctx, vp, (mode_t)vap->va_mode)) != 0) {
return error;
}
if (VATTR_IS_ACTIVE(vap, va_uid) || VATTR_IS_ACTIVE(vap, va_gid)) {
if ((error = mac_vnode_check_setowner(ctx, vp,
VATTR_IS_ACTIVE(vap, va_uid) ? vap->va_uid : -1,
VATTR_IS_ACTIVE(vap, va_gid) ? vap->va_gid : -1))) {
return error;
}
}
if (VATTR_IS_ACTIVE(vap, va_acl) &&
(error = mac_vnode_check_setacl(ctx, vp, vap->va_acl))) {
return error;
}
#endif
/* make sure that the caller is allowed to set this security information */
if (((error = vnode_authattr(vp, vap, &action, ctx)) != 0) ||
((error = vnode_authorize(vp, NULL, action, ctx)) != 0)) {
if (error == EACCES) {
error = EPERM;
}
return error;
}
if ((error = vnode_setattr(vp, vap, ctx)) != 0) {
return error;
}
#if CONFIG_MACF
if (VATTR_IS_ACTIVE(vap, va_mode)) {
mac_vnode_notify_setmode(ctx, vp, (mode_t)vap->va_mode);
}
if (VATTR_IS_ACTIVE(vap, va_uid) || VATTR_IS_ACTIVE(vap, va_gid)) {
mac_vnode_notify_setowner(ctx, vp,
VATTR_IS_ACTIVE(vap, va_uid) ? vap->va_uid : -1,
VATTR_IS_ACTIVE(vap, va_gid) ? vap->va_gid : -1);
}
if (VATTR_IS_ACTIVE(vap, va_acl)) {
mac_vnode_notify_setacl(ctx, vp, vap->va_acl);
}
#endif
return error;
}
/*
* Change mode of a file given a path name.
*
* Returns: 0 Success
* namei:??? [anything namei can return]
* chmod_vnode:??? [anything chmod_vnode can return]
*/
static int
chmodat(vfs_context_t ctx, user_addr_t path, struct vnode_attr *vap,
int fd, int flag, enum uio_seg segflg)
{
struct nameidata nd;
int follow, error;
uint32_t wantparent = 0;
#if CONFIG_FILE_LEASES
wantparent = WANTPARENT;
#endif
follow = (flag & (AT_SYMLINK_NOFOLLOW | AT_SYMLINK_NOFOLLOW_ANY)) ? NOFOLLOW : FOLLOW;
NDINIT(&nd, LOOKUP, OP_SETATTR, follow | AUDITVNPATH1 | wantparent,
segflg, path, ctx);
if (flag & AT_SYMLINK_NOFOLLOW_ANY) {
nd.ni_flag |= NAMEI_NOFOLLOW_ANY;
}
if ((error = nameiat(&nd, fd))) {
return error;
}
#if CONFIG_FILE_LEASES
vnode_breakdirlease(nd.ni_dvp, false, O_WRONLY);
vnode_put(nd.ni_dvp);
#endif
error = chmod_vnode(ctx, nd.ni_vp, vap);
vnode_put(nd.ni_vp);
nameidone(&nd);
return error;
}
static int
chmod_extended_init(struct vnode_attr *pva, kauth_filesec_t *pxsecdst, int mode, uid_t uid,
gid_t gid, user_addr_t xsecurity)
{
int error;
VATTR_INIT(pva);
if (mode != -1) {
VATTR_SET(pva, va_mode, mode & ALLPERMS);
} else {
pva->va_mode = 0;
}
if (uid != KAUTH_UID_NONE) {
VATTR_SET(pva, va_uid, uid);
}
if (gid != KAUTH_GID_NONE) {
VATTR_SET(pva, va_gid, gid);
}
*pxsecdst = NULL;
switch (xsecurity) {
case USER_ADDR_NULL:
break;
case CAST_USER_ADDR_T((void *)1): /* _FILESEC_REMOVE_ACL */
VATTR_SET(pva, va_acl, NULL);
break;
default:
if ((error = kauth_copyinfilesec(xsecurity, pxsecdst)) != 0) {
return error;
}
VATTR_SET(pva, va_acl, &(*pxsecdst)->fsec_acl);
pva->va_vaflags |= VA_FILESEC_ACL;
KAUTH_DEBUG("CHMOD - setting ACL with %d entries", pva->va_acl->acl_entrycount);
break;
}
return 0;
}
/*
* chmod_extended: Change the mode of a file given a path name; with extended
* argument list (including extended security (ACL)).
*
* Parameters: p Process requesting the open
* uap User argument descriptor (see below)
* retval (ignored)
*
* Indirect: uap->path Path to object (same as 'chmod')
* uap->uid UID to set
* uap->gid GID to set
* uap->mode File mode to set (same as 'chmod')
* uap->xsecurity ACL to set (or delete)
*
* Returns: 0 Success
* !0 errno value
*
* Notes: The kauth_filesec_t in 'va', if any, is in host byte order.
*
* XXX: We should enummerate the possible errno values here, and where
* in the code they originated.
*/
int
chmod_extended(__unused proc_t p, struct chmod_extended_args *uap, __unused int32_t *retval)
{
int error;
struct vnode_attr va;
kauth_filesec_t xsecdst = NULL;
AUDIT_ARG(owner, uap->uid, uap->gid);
error = chmod_extended_init(&va, &xsecdst, uap->mode, uap->uid,
uap->gid, uap->xsecurity);
if (error) {
return error;
}
error = chmodat(vfs_context_current(), uap->path, &va, AT_FDCWD, 0,
UIO_USERSPACE);
if (xsecdst != NULL) {
kauth_filesec_free(xsecdst);
}
return error;
}
/*
* Returns: 0 Success
* chmodat:??? [anything chmodat can return]
*/
static int
fchmodat_internal(vfs_context_t ctx, user_addr_t path, int mode, int fd,
int flag, enum uio_seg segflg)
{
struct vnode_attr va;
VATTR_INIT(&va);
VATTR_SET(&va, va_mode, mode & ALLPERMS);
return chmodat(ctx, path, &va, fd, flag, segflg);
}
int
chmod(__unused proc_t p, struct chmod_args *uap, __unused int32_t *retval)
{
return fchmodat_internal(vfs_context_current(), uap->path, uap->mode,
AT_FDCWD, 0, UIO_USERSPACE);
}
int
fchmodat(__unused proc_t p, struct fchmodat_args *uap, __unused int32_t *retval)
{
if (uap->flag & ~(AT_SYMLINK_NOFOLLOW | AT_SYMLINK_NOFOLLOW_ANY)) {
return EINVAL;
}
return fchmodat_internal(vfs_context_current(), uap->path, uap->mode,
uap->fd, uap->flag, UIO_USERSPACE);
}
/*
* Change mode of a file given a file descriptor.
*/
static int
fchmod1(__unused proc_t p, int fd, struct vnode_attr *vap)
{
vnode_t vp;
int error;
AUDIT_ARG(fd, fd);
if ((error = file_vnode(fd, &vp)) != 0) {
return error;
}
if ((error = vnode_getwithref(vp)) != 0) {
file_drop(fd);
return error;
}
AUDIT_ARG(vnpath, vp, ARG_VNODE1);
#if CONFIG_FILE_LEASES
vnode_breakdirlease(vp, true, O_WRONLY);
#endif
error = chmod_vnode(vfs_context_current(), vp, vap);
(void)vnode_put(vp);
file_drop(fd);
return error;
}
/*
* fchmod_extended: Change mode of a file given a file descriptor; with
* extended argument list (including extended security (ACL)).
*
* Parameters: p Process requesting to change file mode
* uap User argument descriptor (see below)
* retval (ignored)
*
* Indirect: uap->mode File mode to set (same as 'chmod')
* uap->uid UID to set
* uap->gid GID to set
* uap->xsecurity ACL to set (or delete)
* uap->fd File descriptor of file to change mode
*
* Returns: 0 Success
* !0 errno value
*
*/
int
fchmod_extended(proc_t p, struct fchmod_extended_args *uap, __unused int32_t *retval)
{
int error;
struct vnode_attr va;
kauth_filesec_t xsecdst = NULL;
AUDIT_ARG(owner, uap->uid, uap->gid);
error = chmod_extended_init(&va, &xsecdst, uap->mode, uap->uid,
uap->gid, uap->xsecurity);
if (error) {
return error;
}
error = fchmod1(p, uap->fd, &va);
if (xsecdst != NULL) {
kauth_filesec_free(xsecdst);
}
return error;
}
int
fchmod(proc_t p, struct fchmod_args *uap, __unused int32_t *retval)
{
struct vnode_attr va;
VATTR_INIT(&va);
VATTR_SET(&va, va_mode, uap->mode & ALLPERMS);
return fchmod1(p, uap->fd, &va);
}
static int
vn_chown_internal(__unused vfs_context_t ctx, vnode_t vp, uid_t uid, gid_t gid)
{
struct vnode_attr va;
kauth_action_t action;
int error;
VATTR_INIT(&va);
if (uid != (uid_t)VNOVAL) {
VATTR_SET(&va, va_uid, uid);
}
if (gid != (gid_t)VNOVAL) {
VATTR_SET(&va, va_gid, gid);
}
#if NAMEDSTREAMS
/* chown calls are not allowed for resource forks. */
if (vp->v_flag & VISNAMEDSTREAM) {
error = EPERM;
goto out;
}
#endif
#if CONFIG_MACF
error = mac_vnode_check_setowner(ctx, vp, uid, gid);
if (error) {
goto out;
}
#endif
/* preflight and authorize attribute changes */
if ((error = vnode_authattr(vp, &va, &action, ctx)) != 0) {
goto out;
}
if (action && ((error = vnode_authorize(vp, NULL, action, ctx)) != 0)) {
/*
* EACCES is only allowed from namei(); permissions failure should
* return EPERM, so we need to translate the error code.
*/
if (error == EACCES) {
error = EPERM;
}
goto out;
}
#if CONFIG_FILE_LEASES
vnode_breakdirlease(vp, true, O_WRONLY);
#endif
error = vnode_setattr(vp, &va, ctx);
#if CONFIG_MACF
if (error == 0) {
mac_vnode_notify_setowner(ctx, vp, uid, gid);
}
#endif
out:
return error;
}
/*
* Set ownership given a path name.
*/
/* ARGSUSED */
static int
fchownat_internal(vfs_context_t ctx, int fd, user_addr_t path, uid_t uid,
gid_t gid, int flag, enum uio_seg segflg)
{
vnode_t vp;
int error;
struct nameidata nd;
int follow;
AUDIT_ARG(owner, uid, gid);
follow = (flag & (AT_SYMLINK_NOFOLLOW | AT_SYMLINK_NOFOLLOW_ANY)) ? NOFOLLOW : FOLLOW;
NDINIT(&nd, LOOKUP, OP_SETATTR, follow | AUDITVNPATH1, segflg, path, ctx);
if (flag & AT_SYMLINK_NOFOLLOW_ANY) {
nd.ni_flag |= NAMEI_NOFOLLOW_ANY;
}
error = nameiat(&nd, fd);
if (error) {
return error;
}
vp = nd.ni_vp;
error = vn_chown_internal(ctx, vp, uid, gid);
nameidone(&nd);
vnode_put(vp);
return error;
}
int
chown(__unused proc_t p, struct chown_args *uap, __unused int32_t *retval)
{
return fchownat_internal(vfs_context_current(), AT_FDCWD, uap->path,
uap->uid, uap->gid, 0, UIO_USERSPACE);
}
int
lchown(__unused proc_t p, struct lchown_args *uap, __unused int32_t *retval)
{
return fchownat_internal(vfs_context_current(), AT_FDCWD, uap->path,
uap->owner, uap->group, AT_SYMLINK_NOFOLLOW, UIO_USERSPACE);
}
int
fchownat(__unused proc_t p, struct fchownat_args *uap, __unused int32_t *retval)
{
if (uap->flag & ~AT_SYMLINK_NOFOLLOW) {
return EINVAL;
}
return fchownat_internal(vfs_context_current(), uap->fd, uap->path,
uap->uid, uap->gid, uap->flag, UIO_USERSPACE);
}
/*
* Set ownership given a file descriptor.
*/
/* ARGSUSED */
int
fchown(__unused proc_t p, struct fchown_args *uap, __unused int32_t *retval)
{
vfs_context_t ctx = vfs_context_current();
vnode_t vp;
int error;
AUDIT_ARG(owner, uap->uid, uap->gid);
AUDIT_ARG(fd, uap->fd);
if ((error = file_vnode(uap->fd, &vp))) {
return error;
}
if ((error = vnode_getwithref(vp))) {
file_drop(uap->fd);
return error;
}
AUDIT_ARG(vnpath, vp, ARG_VNODE1);
error = vn_chown_internal(ctx, vp, uap->uid, uap->gid);
(void)vnode_put(vp);
file_drop(uap->fd);
return error;
}
static int
getutimes(user_addr_t usrtvp, struct timespec *tsp)
{
int error;
if (usrtvp == USER_ADDR_NULL) {
struct timeval old_tv;
/* XXX Y2038 bug because of microtime argument */
microtime(&old_tv);
TIMEVAL_TO_TIMESPEC(&old_tv, &tsp[0]);
tsp[1] = tsp[0];
} else {
if (IS_64BIT_PROCESS(current_proc())) {
struct user64_timeval tv[2];
error = copyin(usrtvp, (void *)tv, sizeof(tv));
if (error) {
return error;
}
TIMEVAL64_TO_TIMESPEC(&tv[0], &tsp[0]);
TIMEVAL64_TO_TIMESPEC(&tv[1], &tsp[1]);
} else {
struct user32_timeval tv[2];
error = copyin(usrtvp, (void *)tv, sizeof(tv));
if (error) {
return error;
}
TIMEVAL_TO_TIMESPEC(&tv[0], &tsp[0]);
TIMEVAL_TO_TIMESPEC(&tv[1], &tsp[1]);
}
}
return 0;
}
static int
setutimes(vfs_context_t ctx, vnode_t vp, const struct timespec *ts,
int nullflag)
{
int error;
struct vnode_attr va;
kauth_action_t action;
AUDIT_ARG(vnpath, vp, ARG_VNODE1);
VATTR_INIT(&va);
VATTR_SET(&va, va_access_time, ts[0]);
VATTR_SET(&va, va_modify_time, ts[1]);
if (nullflag) {
va.va_vaflags |= VA_UTIMES_NULL;
}
#if NAMEDSTREAMS
/* utimes calls are not allowed for resource forks. */
if (vp->v_flag & VISNAMEDSTREAM) {
error = EPERM;
goto out;
}
#endif
#if CONFIG_MACF
error = mac_vnode_check_setutimes(ctx, vp, ts[0], ts[1]);
if (error) {
goto out;
}
#endif
if ((error = vnode_authattr(vp, &va, &action, ctx)) != 0) {
if (!nullflag && error == EACCES) {
error = EPERM;
}
goto out;
}
/* since we may not need to auth anything, check here */
if ((action != 0) && ((error = vnode_authorize(vp, NULL, action, ctx)) != 0)) {
if (!nullflag && error == EACCES) {
error = EPERM;
}
goto out;
}
error = vnode_setattr(vp, &va, ctx);
#if CONFIG_MACF
if (error == 0) {
mac_vnode_notify_setutimes(ctx, vp, ts[0], ts[1]);
}
#endif
out:
return error;
}
/*
* Set the access and modification times of a file.
*/
/* ARGSUSED */
int
utimes(__unused proc_t p, struct utimes_args *uap, __unused int32_t *retval)
{
struct timespec ts[2];
user_addr_t usrtvp;
int error;
struct nameidata nd;
vfs_context_t ctx = vfs_context_current();
uint32_t wantparent = 0;
#if CONFIG_FILE_LEASES
wantparent = WANTPARENT;
#endif
/*
* AUDIT: Needed to change the order of operations to do the
* name lookup first because auditing wants the path.
*/
NDINIT(&nd, LOOKUP, OP_SETATTR, FOLLOW | AUDITVNPATH1 | wantparent,
UIO_USERSPACE, uap->path, ctx);
error = namei(&nd);
if (error) {
return error;
}
/*
* Fetch the user-supplied time. If usrtvp is USER_ADDR_NULL, we fetch
* the current time instead.
*/
usrtvp = uap->tptr;
if ((error = getutimes(usrtvp, ts)) != 0) {
goto out;
}
#if CONFIG_FILE_LEASES
vnode_breakdirlease(nd.ni_dvp, false, O_WRONLY);
#endif
error = setutimes(ctx, nd.ni_vp, ts, usrtvp == USER_ADDR_NULL);
out:
#if CONFIG_FILE_LEASES
vnode_put(nd.ni_dvp);
#endif
nameidone(&nd);
vnode_put(nd.ni_vp);
return error;
}
/*
* Set the access and modification times of a file.
*/
/* ARGSUSED */
int
futimes(__unused proc_t p, struct futimes_args *uap, __unused int32_t *retval)
{
struct timespec ts[2];
vnode_t vp;
user_addr_t usrtvp;
int error;
AUDIT_ARG(fd, uap->fd);
usrtvp = uap->tptr;
if ((error = getutimes(usrtvp, ts)) != 0) {
return error;
}
if ((error = file_vnode(uap->fd, &vp)) != 0) {
return error;
}
if ((error = vnode_getwithref(vp))) {
file_drop(uap->fd);
return error;
}
#if CONFIG_FILE_LEASES
vnode_breakdirlease(vp, true, O_WRONLY);
#endif
error = setutimes(vfs_context_current(), vp, ts, usrtvp == 0);
vnode_put(vp);
file_drop(uap->fd);
return error;
}
static int
truncate_validate_common(proc_t p, off_t length)
{
rlim_t fsize_limit;
if (length < 0) {
return EINVAL;
}
fsize_limit = proc_limitgetcur(p, RLIMIT_FSIZE);
if ((rlim_t)length > fsize_limit) {
psignal(p, SIGXFSZ);
return EFBIG;
}
return 0;
}
static int
truncate_internal(vnode_t vp, off_t length, kauth_cred_t cred,
vfs_context_t ctx, boolean_t need_auth)
{
struct vnode_attr va;
kauth_action_t action;
int error;
VATTR_INIT(&va);
VATTR_SET(&va, va_data_size, length);
#if CONFIG_MACF
error = mac_vnode_check_truncate(ctx, cred, vp);
if (error) {
return error;
}
#endif
/*
* If we reached here from `ftruncate` then we already did an effective
* `vnode_authorize` upon open. We honour the result from then.
*/
if (need_auth) {
if ((error = vnode_authattr(vp, &va, &action, ctx)) != 0) {
return error;
}
if ((action != 0) && ((error = vnode_authorize(vp, NULL, action, ctx)) != 0)) {
return error;
}
}
#if CONFIG_FILE_LEASES
/* Check if there is a lease placed on the parent directory. */
vnode_breakdirlease(vp, true, O_WRONLY);
/* Now check if there is a lease placed on the file itself. */
(void)vnode_breaklease(vp, O_WRONLY, ctx);
#endif
error = vnode_setattr(vp, &va, ctx);
#if CONFIG_MACF
if (error == 0) {
mac_vnode_notify_truncate(ctx, cred, vp);
}
#endif
return error;
}
/*
* Truncate a file given its path name.
*/
/* ARGSUSED */
int
truncate(proc_t p, struct truncate_args *uap, __unused int32_t *retval)
{
vfs_context_t ctx = vfs_context_current();
vnode_t vp;
int error;
struct nameidata nd;
if ((error = truncate_validate_common(p, uap->length))) {
return error;
}
NDINIT(&nd, LOOKUP, OP_TRUNCATE, FOLLOW | AUDITVNPATH1,
UIO_USERSPACE, uap->path, ctx);
if ((error = namei(&nd))) {
return error;
}
vp = nd.ni_vp;
nameidone(&nd);
error = truncate_internal(vp, uap->length, NOCRED, ctx, true);
vnode_put(vp);
return error;
}
/*
* Truncate a file given a file descriptor.
*/
/* ARGSUSED */
int
ftruncate(proc_t p, struct ftruncate_args *uap, int32_t *retval)
{
vnode_t vp;
struct fileproc *fp;
int error;
AUDIT_ARG(fd, uap->fd);
if ((error = truncate_validate_common(p, uap->length))) {
return error;
}
if ((error = fp_lookup(p, uap->fd, &fp, 0))) {
return error;
}
switch (FILEGLOB_DTYPE(fp->fp_glob)) {
case DTYPE_PSXSHM:
error = pshm_truncate(p, fp, uap->fd, uap->length, retval);
goto out;
case DTYPE_VNODE:
break;
default:
error = EINVAL;
goto out;
}
vp = (vnode_t)fp_get_data(fp);
if ((fp->fp_glob->fg_flag & FWRITE) == 0) {
AUDIT_ARG(vnpath_withref, vp, ARG_VNODE1);
error = EINVAL;
goto out;
}
if ((error = vnode_getwithref(vp)) != 0) {
goto out;
}
AUDIT_ARG(vnpath, vp, ARG_VNODE1);
error = truncate_internal(vp, uap->length, fp->fp_glob->fg_cred,
vfs_context_current(), false);
vnode_put(vp);
out:
file_drop(uap->fd);
return error;
}
/*
* Sync an open file with synchronized I/O _file_ integrity completion
*/
/* ARGSUSED */
int
fsync(proc_t p, struct fsync_args *uap, __unused int32_t *retval)
{
__pthread_testcancel(1);
return fsync_common(p, uap, MNT_WAIT);
}
/*
* Sync an open file with synchronized I/O _file_ integrity completion
*
* Notes: This is a legacy support function that does not test for
* thread cancellation points.
*/
/* ARGSUSED */
int
fsync_nocancel(proc_t p, struct fsync_nocancel_args *uap, __unused int32_t *retval)
{
return fsync_common(p, (struct fsync_args *)uap, MNT_WAIT);
}
/*
* Sync an open file with synchronized I/O _data_ integrity completion
*/
/* ARGSUSED */
int
fdatasync(proc_t p, struct fdatasync_args *uap, __unused int32_t *retval)
{
__pthread_testcancel(1);
return fsync_common(p, (struct fsync_args *)uap, MNT_DWAIT);
}
/*
* fsync_common
*
* Common fsync code to support both synchronized I/O file integrity completion
* (normal fsync) and synchronized I/O data integrity completion (fdatasync).
*
* If 'flags' is MNT_DWAIT, the caller is requesting data integrity, which
* will only guarantee that the file data contents are retrievable. If
* 'flags' is MNT_WAIT, the caller is rewuesting file integrity, which also
* includes additional metadata unnecessary for retrieving the file data
* contents, such as atime, mtime, ctime, etc., also be committed to stable
* storage.
*
* Parameters: p The process
* uap->fd The descriptor to synchronize
* flags The data integrity flags
*
* Returns: int Success
* fp_getfvp:EBADF Bad file descriptor
* fp_getfvp:ENOTSUP fd does not refer to a vnode
* VNOP_FSYNC:??? unspecified
*
* Notes: We use struct fsync_args because it is a short name, and all
* caller argument structures are otherwise identical.
*/
static int
fsync_common(proc_t p, struct fsync_args *uap, int flags)
{
vnode_t vp;
struct fileproc *fp;
vfs_context_t ctx = vfs_context_current();
int error;
AUDIT_ARG(fd, uap->fd);
if ((error = fp_getfvp(p, uap->fd, &fp, &vp))) {
return error;
}
if ((error = vnode_getwithref(vp))) {
file_drop(uap->fd);
return error;
}
AUDIT_ARG(vnpath, vp, ARG_VNODE1);
error = VNOP_FSYNC(vp, flags, ctx);
#if NAMEDRSRCFORK
/* Sync resource fork shadow file if necessary. */
if ((error == 0) &&
(vp->v_flag & VISNAMEDSTREAM) &&
(vp->v_parent != NULLVP) &&
vnode_isshadow(vp) &&
(fp->fp_glob->fg_flag & FWASWRITTEN)) {
(void) vnode_flushnamedstream(vp->v_parent, vp, ctx);
}
#endif
(void)vnode_put(vp);
file_drop(uap->fd);
return error;
}
/*
* Duplicate files. Source must be a file, target must be a file or
* must not exist.
*
* XXX Copyfile authorisation checking is woefully inadequate, and will not
* perform inheritance correctly.
*/
/* ARGSUSED */
int
copyfile(__unused proc_t p, struct copyfile_args *uap, __unused int32_t *retval)
{
vnode_t tvp, fvp, tdvp, sdvp;
struct nameidata fromnd, tond;
int error;
vfs_context_t ctx = vfs_context_current();
/* Check that the flags are valid. */
if (uap->flags & ~CPF_MASK) {
return EINVAL;
}
NDINIT(&fromnd, LOOKUP, OP_COPYFILE, AUDITVNPATH1,
UIO_USERSPACE, uap->from, ctx);
if ((error = namei(&fromnd))) {
return error;
}
fvp = fromnd.ni_vp;
NDINIT(&tond, CREATE, OP_LINK,
LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | AUDITVNPATH2 | CN_NBMOUNTLOOK,
UIO_USERSPACE, uap->to, ctx);
if ((error = namei(&tond))) {
goto out1;
}
tdvp = tond.ni_dvp;
tvp = tond.ni_vp;
if (tvp != NULL) {
if (!(uap->flags & CPF_OVERWRITE)) {
error = EEXIST;
goto out;
}
}
if (fvp->v_type == VDIR || (tvp && tvp->v_type == VDIR)) {
error = EISDIR;
goto out;
}
if (fvp->v_type == VSOCK && fvp->v_tag != VT_FDESC) {
error = EOPNOTSUPP;
goto out;
}
#if CONFIG_MACF
if ((error = mac_vnode_check_copyfile(ctx, tdvp, tvp, fvp, &tond.ni_cnd, (mode_t)uap->mode, uap->flags)) != 0) {
goto out;
}
#endif /* CONFIG_MACF */
if ((error = vnode_authorize(fvp, NULL, KAUTH_VNODE_READ_DATA, ctx)) != 0) {
goto out;
}
if (tvp) {
if ((error = vnode_authorize(tvp, tdvp, KAUTH_VNODE_DELETE, ctx)) != 0) {
goto out;
}
}
if ((error = vnode_authorize(tdvp, NULL, KAUTH_VNODE_ADD_FILE, ctx)) != 0) {
goto out;
}
if (fvp == tdvp) {
error = EINVAL;
}
/*
* If source is the same as the destination (that is the
* same inode number) then there is nothing to do.
* (fixed to have POSIX semantics - CSM 3/2/98)
*/
if (fvp == tvp) {
error = -1;
}
#if CONFIG_FILE_LEASES
vnode_breakdirlease(tdvp, false, O_WRONLY);
#endif
if (!error) {
error = VNOP_COPYFILE(fvp, tdvp, tvp, &tond.ni_cnd, uap->mode, uap->flags, ctx);
}
out:
sdvp = tond.ni_startdir;
/*
* nameidone has to happen before we vnode_put(tdvp)
* since it may need to release the fs_nodelock on the tdvp
*/
nameidone(&tond);
if (tvp) {
vnode_put(tvp);
}
vnode_put(tdvp);
vnode_put(sdvp);
out1:
vnode_put(fvp);
nameidone(&fromnd);
if (error == -1) {
return 0;
}
return error;
}
#define CLONE_SNAPSHOT_FALLBACKS_ENABLED 1
/*
* Helper function for doing clones. The caller is expected to provide an
* iocounted source vnode and release it.
*/
static int
clonefile_internal(vnode_t fvp, boolean_t data_read_authorised, int dst_dirfd,
user_addr_t dst, uint32_t flags, vfs_context_t ctx)
{
vnode_t tvp, tdvp;
struct nameidata tond;
int error;
int follow;
boolean_t free_src_acl;
boolean_t attr_cleanup;
enum vtype v_type;
kauth_action_t action;
struct componentname *cnp;
uint32_t defaulted = 0;
struct vnode_attr va;
struct vnode_attr nva;
uint32_t vnop_flags;
v_type = vnode_vtype(fvp);
switch (v_type) {
case VLNK:
/* FALLTHRU */
case VREG:
action = KAUTH_VNODE_ADD_FILE;
break;
case VDIR:
if (vnode_isvroot(fvp) || vnode_ismount(fvp) ||
fvp->v_mountedhere) {
return EINVAL;
}
action = KAUTH_VNODE_ADD_SUBDIRECTORY;
break;
default:
return EINVAL;
}
AUDIT_ARG(fd2, dst_dirfd);
AUDIT_ARG(value32, flags);
follow = (flags & CLONE_NOFOLLOW) ? NOFOLLOW : FOLLOW;
NDINIT(&tond, CREATE, OP_LINK, follow | WANTPARENT | AUDITVNPATH2,
UIO_USERSPACE, dst, ctx);
if ((error = nameiat(&tond, dst_dirfd))) {
return error;
}
cnp = &tond.ni_cnd;
tdvp = tond.ni_dvp;
tvp = tond.ni_vp;
free_src_acl = FALSE;
attr_cleanup = FALSE;
if (tvp != NULL) {
error = EEXIST;
goto out;
}
if (vnode_mount(tdvp) != vnode_mount(fvp)) {
error = EXDEV;
goto out;
}
#if CONFIG_MACF
if ((error = mac_vnode_check_clone(ctx, tdvp, fvp, cnp))) {
goto out;
}
#endif
if ((error = vnode_authorize(tdvp, NULL, action, ctx))) {
goto out;
}
action = KAUTH_VNODE_GENERIC_READ_BITS;
if (data_read_authorised) {
action &= ~KAUTH_VNODE_READ_DATA;
}
if ((error = vnode_authorize(fvp, NULL, action, ctx))) {
goto out;
}
/*
* certain attributes may need to be changed from the source, we ask for
* those here with the exception of source file's ACLs unless the CLONE_ACL
* flag is specified. By default, the clone file will inherit the target
* directory's ACLs unless the the CLONE_ACL flag is specified then it
* will inherit the source file's ACLs instead.
*/
VATTR_INIT(&va);
VATTR_WANTED(&va, va_uid);
VATTR_WANTED(&va, va_gid);
VATTR_WANTED(&va, va_mode);
VATTR_WANTED(&va, va_flags);
if (flags & CLONE_ACL) {
VATTR_WANTED(&va, va_acl);
}
if ((error = vnode_getattr(fvp, &va, ctx)) != 0) {
goto out;
}
VATTR_INIT(&nva);
VATTR_SET(&nva, va_type, v_type);
if (VATTR_IS_SUPPORTED(&va, va_acl) && va.va_acl != NULL) {
VATTR_SET(&nva, va_acl, va.va_acl);
free_src_acl = TRUE;
}
/* Handle ACL inheritance, initialize vap. */
if (v_type == VLNK) {
error = vnode_authattr_new(tdvp, &nva, 0, ctx);
} else {
error = vn_attribute_prepare(tdvp, &nva, &defaulted, ctx);
if (error) {
goto out;
}
attr_cleanup = TRUE;
}
vnop_flags = VNODE_CLONEFILE_DEFAULT;
/*
* We've got initial values for all security parameters,
* If we are superuser, then we can change owners to be the
* same as the source. Both superuser and the owner have default
* WRITE_SECURITY privileges so all other fields can be taken
* from source as well.
*/
if (!(flags & CLONE_NOOWNERCOPY) && vfs_context_issuser(ctx)) {
if (VATTR_IS_SUPPORTED(&va, va_uid)) {
VATTR_SET(&nva, va_uid, va.va_uid);
}
if (VATTR_IS_SUPPORTED(&va, va_gid)) {
VATTR_SET(&nva, va_gid, va.va_gid);
}
} else {
vnop_flags |= VNODE_CLONEFILE_NOOWNERCOPY;
}
if (VATTR_IS_SUPPORTED(&va, va_mode)) {
VATTR_SET(&nva, va_mode, va.va_mode);
}
if (VATTR_IS_SUPPORTED(&va, va_flags)) {
VATTR_SET(&nva, va_flags,
((va.va_flags & ~(UF_DATAVAULT | SF_RESTRICTED)) | /* Turn off from source */
(nva.va_flags & (UF_DATAVAULT | SF_RESTRICTED))));
}
#if CONFIG_FILE_LEASES
vnode_breakdirlease(tdvp, false, O_WRONLY);
#endif
error = VNOP_CLONEFILE(fvp, tdvp, &tvp, cnp, &nva, vnop_flags, ctx);
if (!error && tvp) {
int update_flags = 0;
#if CONFIG_FSE
int fsevent;
#endif /* CONFIG_FSE */
/*
* If some of the requested attributes weren't handled by the
* VNOP, use our fallback code.
*/
if (!VATTR_ALL_SUPPORTED(&nva)) {
(void)vnode_setattr_fallback(tvp, &nva, ctx);
}
#if CONFIG_MACF
(void)vnode_label(vnode_mount(tvp), tdvp, tvp, cnp,
VNODE_LABEL_CREATE, ctx);
#endif
// Make sure the name & parent pointers are hooked up
if (tvp->v_name == NULL) {
update_flags |= VNODE_UPDATE_NAME;
}
if (tvp->v_parent == NULLVP) {
update_flags |= VNODE_UPDATE_PARENT;
}
if (update_flags) {
(void)vnode_update_identity(tvp, tdvp, cnp->cn_nameptr,
cnp->cn_namelen, cnp->cn_hash, update_flags);
}
#if CONFIG_FSE
switch (vnode_vtype(tvp)) {
case VLNK:
/* FALLTHRU */
case VREG:
fsevent = FSE_CREATE_FILE;
break;
case VDIR:
fsevent = FSE_CREATE_DIR;
break;
default:
goto out;
}
if (need_fsevent(fsevent, tvp)) {
/*
* The following is a sequence of three explicit events.
* A pair of FSE_CLONE events representing the source and destination
* followed by an FSE_CREATE_[FILE | DIR] for the destination.
* fseventsd may coalesce the destination clone and create events
* into a single event resulting in the following sequence for a client
* FSE_CLONE (src)
* FSE_CLONE | FSE_CREATE (dst)
*/
add_fsevent(FSE_CLONE, ctx, FSE_ARG_VNODE, fvp, FSE_ARG_VNODE, tvp,
FSE_ARG_DONE);
add_fsevent(fsevent, ctx, FSE_ARG_VNODE, tvp,
FSE_ARG_DONE);
}
#endif /* CONFIG_FSE */
}
out:
if (attr_cleanup) {
vn_attribute_cleanup(&nva, defaulted);
}
if (free_src_acl && va.va_acl) {
kauth_acl_free(va.va_acl);
}
nameidone(&tond);
if (tvp) {
vnode_put(tvp);
}
vnode_put(tdvp);
return error;
}
/*
* clone files or directories, target must not exist.
*/
/* ARGSUSED */
int
clonefileat(__unused proc_t p, struct clonefileat_args *uap,
__unused int32_t *retval)
{
vnode_t fvp;
struct nameidata fromnd;
int follow;
int error;
vfs_context_t ctx = vfs_context_current();
/* Check that the flags are valid. */
if (uap->flags & ~(CLONE_NOFOLLOW | CLONE_NOOWNERCOPY | CLONE_ACL)) {
return EINVAL;
}
AUDIT_ARG(fd, uap->src_dirfd);
follow = (uap->flags & CLONE_NOFOLLOW) ? NOFOLLOW : FOLLOW;
NDINIT(&fromnd, LOOKUP, OP_COPYFILE, follow | AUDITVNPATH1,
UIO_USERSPACE, uap->src, ctx);
if ((error = nameiat(&fromnd, uap->src_dirfd))) {
return error;
}
fvp = fromnd.ni_vp;
nameidone(&fromnd);
error = clonefile_internal(fvp, FALSE, uap->dst_dirfd, uap->dst,
uap->flags, ctx);
vnode_put(fvp);
return error;
}
int
fclonefileat(__unused proc_t p, struct fclonefileat_args *uap,
__unused int32_t *retval)
{
vnode_t fvp;
struct fileproc *fp;
int error;
vfs_context_t ctx = vfs_context_current();
/* Check that the flags are valid. */
if (uap->flags & ~(CLONE_NOFOLLOW | CLONE_NOOWNERCOPY | CLONE_ACL)) {
return EINVAL;
}
AUDIT_ARG(fd, uap->src_fd);
error = fp_getfvp(p, uap->src_fd, &fp, &fvp);
if (error) {
return error;
}
if ((fp->fp_glob->fg_flag & FREAD) == 0) {
AUDIT_ARG(vnpath_withref, fvp, ARG_VNODE1);
error = EBADF;
goto out;
}
if ((error = vnode_getwithref(fvp))) {
goto out;
}
AUDIT_ARG(vnpath, fvp, ARG_VNODE1);
error = clonefile_internal(fvp, TRUE, uap->dst_dirfd, uap->dst,
uap->flags, ctx);
vnode_put(fvp);
out:
file_drop(uap->src_fd);
return error;
}
static int
rename_submounts_callback(mount_t mp, void *arg)
{
int error = 0;
mount_t pmp = (mount_t)arg;
int prefix_len = (int)strlen(pmp->mnt_vfsstat.f_mntonname);
if (strncmp(mp->mnt_vfsstat.f_mntonname, pmp->mnt_vfsstat.f_mntonname, prefix_len) != 0) {
return 0;
}
if (mp->mnt_vfsstat.f_mntonname[prefix_len] != '/') {
return 0;
}
if ((error = vfs_busy(mp, LK_NOWAIT))) {
printf("vfs_busy failed with %d for %s\n", error, mp->mnt_vfsstat.f_mntonname);
return -1;
}
size_t pathlen = MAXPATHLEN;
if ((error = vn_getpath_ext(mp->mnt_vnodecovered, NULL, mp->mnt_vfsstat.f_mntonname, &pathlen, VN_GETPATH_FSENTER))) {
printf("vn_getpath_ext failed with %d for mnt_vnodecovered of %s\n", error, mp->mnt_vfsstat.f_mntonname);
}
vfs_unbusy(mp);
return error;
}
/*
* Rename files. Source and destination must either both be directories,
* or both not be directories. If target is a directory, it must be empty.
*/
/* ARGSUSED */
static int
renameat_internal(vfs_context_t ctx, int fromfd, user_addr_t from,
int tofd, user_addr_t to, int segflg, u_int uflags)
{
vnode_t tvp, tdvp;
vnode_t fvp, fdvp;
vnode_t mnt_fvp;
struct nameidata *fromnd, *tond;
int error = 0;
int do_retry;
int retry_count;
int mntrename;
int need_event;
int need_kpath2;
int has_listeners;
const char *oname = NULL;
char *from_name = NULL, *to_name = NULL;
char *from_name_no_firmlink = NULL, *to_name_no_firmlink = NULL;
int from_len = 0, to_len = 0;
int from_len_no_firmlink = 0, to_len_no_firmlink = 0;
int holding_mntlock;
int vn_authorize_skipped;
mount_t locked_mp = NULL;
vnode_t oparent = NULLVP;
#if CONFIG_FSE
fse_info from_finfo = {}, to_finfo;
#endif
int from_truncated = 0, to_truncated = 0;
int from_truncated_no_firmlink = 0, to_truncated_no_firmlink = 0;
int batched = 0;
struct vnode_attr *fvap, *tvap;
int continuing = 0;
vfs_rename_flags_t flags = uflags & VFS_RENAME_FLAGS_MASK;
int32_t nofollow_any = 0;
/* carving out a chunk for structs that are too big to be on stack. */
struct {
struct nameidata from_node, to_node;
struct vnode_attr fv_attr, tv_attr;
} * __rename_data;
__rename_data = kalloc_type(typeof(*__rename_data), Z_WAITOK);
fromnd = &__rename_data->from_node;
tond = &__rename_data->to_node;
holding_mntlock = 0;
do_retry = 0;
retry_count = 0;
retry:
fvp = tvp = NULL;
fdvp = tdvp = NULL;
fvap = tvap = NULL;
mnt_fvp = NULLVP;
mntrename = FALSE;
vn_authorize_skipped = FALSE;
if (uflags & RENAME_NOFOLLOW_ANY) {
nofollow_any = NAMEI_NOFOLLOW_ANY;
}
NDINIT(fromnd, DELETE, OP_UNLINK, WANTPARENT | AUDITVNPATH1,
segflg, from, ctx);
fromnd->ni_flag = NAMEI_COMPOUNDRENAME | nofollow_any;
NDINIT(tond, RENAME, OP_RENAME, WANTPARENT | AUDITVNPATH2 | CN_NBMOUNTLOOK,
segflg, to, ctx);
tond->ni_flag = NAMEI_COMPOUNDRENAME | nofollow_any;
continue_lookup:
if ((fromnd->ni_flag & NAMEI_CONTLOOKUP) != 0 || !continuing) {
if ((error = nameiat(fromnd, fromfd))) {
goto out1;
}
fdvp = fromnd->ni_dvp;
fvp = fromnd->ni_vp;
if (fvp && fvp->v_type == VDIR) {
tond->ni_cnd.cn_flags |= WILLBEDIR;
}
}
if ((tond->ni_flag & NAMEI_CONTLOOKUP) != 0 || !continuing) {
if ((error = nameiat(tond, tofd))) {
/*
* Translate error code for rename("dir1", "dir2/.").
*/
if (error == EISDIR && fvp->v_type == VDIR) {
error = EINVAL;
}
goto out1;
}
tdvp = tond->ni_dvp;
tvp = tond->ni_vp;
}
#if DEVELOPMENT || DEBUG
/*
* XXX VSWAP: Check for entitlements or special flag here
* so we can restrict access appropriately.
*/
#else /* DEVELOPMENT || DEBUG */
if (fromnd->ni_vp && vnode_isswap(fromnd->ni_vp) && (ctx != vfs_context_kernel())) {
error = EPERM;
goto out1;
}
if (tond->ni_vp && vnode_isswap(tond->ni_vp) && (ctx != vfs_context_kernel())) {
error = EPERM;
goto out1;
}
#endif /* DEVELOPMENT || DEBUG */
if (!tvp && ISSET(flags, VFS_RENAME_SWAP)) {
error = ENOENT;
goto out1;
}
if (tvp && ISSET(flags, VFS_RENAME_EXCL)) {
int32_t pval = 0;
int err = 0;
/*
* We allow rename with VFS_RENAME_EXCL flag for an existing file which
* has the same name as target iff the following conditions are met:
* 1. the target file system is case insensitive
* 2. source and target directories are the same
* 3. source and target files are the same
* 4. name only differs in case (determined by underlying filesystem)
*/
if (fvp != tvp || fdvp != tdvp) {
error = EEXIST;
goto out1;
}
/*
* Assume that the target file system is case sensitive if
* _PC_CASE_SENSITIVE selector isn't supported.
*/
err = VNOP_PATHCONF(tvp, _PC_CASE_SENSITIVE, &pval, ctx);
if (err != 0 || pval != 0) {
error = EEXIST;
goto out1;
}
}
batched = vnode_compound_rename_available(fdvp);
#if CONFIG_FSE
need_event = need_fsevent(FSE_RENAME, fdvp);
if (need_event) {
if (fvp) {
get_fse_info(fvp, &from_finfo, ctx);
} else {
error = vfs_get_notify_attributes(&__rename_data->fv_attr);
if (error) {
goto out1;
}
fvap = &__rename_data->fv_attr;
}
if (tvp) {
get_fse_info(tvp, &to_finfo, ctx);
} else if (batched) {
error = vfs_get_notify_attributes(&__rename_data->tv_attr);
if (error) {
goto out1;
}
tvap = &__rename_data->tv_attr;
}
}
#else
need_event = 0;
#endif /* CONFIG_FSE */
has_listeners = kauth_authorize_fileop_has_listeners();
need_kpath2 = 0;
#if CONFIG_AUDIT
if (AUDIT_RECORD_EXISTS()) {
need_kpath2 = 1;
}
#endif
if (need_event || has_listeners) {
if (from_name == NULL) {
GET_PATH(from_name);
}
from_len = safe_getpath(fdvp, fromnd->ni_cnd.cn_nameptr, from_name, MAXPATHLEN, &from_truncated);
if (from_name_no_firmlink == NULL) {
GET_PATH(from_name_no_firmlink);
}
from_len_no_firmlink = safe_getpath_no_firmlink(fdvp, fromnd->ni_cnd.cn_nameptr, from_name_no_firmlink, MAXPATHLEN, &from_truncated_no_firmlink);
}
if (need_event || need_kpath2 || has_listeners) {
if (to_name == NULL) {
GET_PATH(to_name);
}
to_len = safe_getpath(tdvp, tond->ni_cnd.cn_nameptr, to_name, MAXPATHLEN, &to_truncated);
if (to_name_no_firmlink == NULL) {
GET_PATH(to_name_no_firmlink);
}
to_len_no_firmlink = safe_getpath_no_firmlink(tdvp, tond->ni_cnd.cn_nameptr, to_name_no_firmlink, MAXPATHLEN, &to_truncated_no_firmlink);
if (to_name && need_kpath2) {
AUDIT_ARG(kpath, to_name, ARG_KPATH2);
}
}
if (!fvp) {
/*
* Claim: this check will never reject a valid rename.
* For success, either fvp must be on the same mount as tdvp, or fvp must sit atop a vnode on the same mount as tdvp.
* Suppose fdvp and tdvp are not on the same mount.
* If fvp is on the same mount as tdvp, then fvp is not on the same mount as fdvp, so fvp is the root of its filesystem. If fvp is the root,
* then you can't move it to within another dir on the same mountpoint.
* If fvp sits atop a vnode on the same mount as fdvp, then that vnode must be part of the same mount as fdvp, which is a contradiction.
*
* If this check passes, then we are safe to pass these vnodes to the same FS.
*/
if (fdvp->v_mount != tdvp->v_mount) {
error = EXDEV;
goto out1;
}
goto skipped_lookup;
}
/*
* If the source and destination are the same (i.e. they're
* links to the same vnode) and the target file system is
* case sensitive, then there is nothing to do.
*
* XXX Come back to this.
*/
if (fvp == tvp) {
int pathconf_val;
/*
* Note: if _PC_CASE_SENSITIVE selector isn't supported,
* then assume that this file system is case sensitive.
*/
if (VNOP_PATHCONF(fvp, _PC_CASE_SENSITIVE, &pathconf_val, ctx) != 0 ||
pathconf_val != 0) {
vn_authorize_skipped = TRUE;
goto out1;
}
}
/*
* Allow the renaming of mount points.
* - target must not exist
* - target must reside in the same directory as source
* - union mounts cannot be renamed
* - the root fs, and tightly-linked system volumes, cannot be renamed
*
* XXX Handle this in VFS after a continued lookup (if we missed
* in the cache to start off)
*
* N.B. If RENAME_SWAP is being used, then @tvp != NULL and so
* we'll skip past here. The file system is responsible for
* checking that @tvp is not a descendent of @fvp and vice versa
* so it should always return EINVAL if either @tvp or @fvp is the
* root of a volume.
*/
if ((fvp->v_flag & VROOT) &&
(fvp->v_type == VDIR) &&
(tvp == NULL) &&
(fvp->v_mountedhere == NULL) &&
(fdvp == tdvp) &&
((fvp->v_mount->mnt_flag & (MNT_UNION | MNT_ROOTFS)) == 0) &&
((fvp->v_mount->mnt_kern_flag & MNTK_SYSTEM) == 0) &&
(fvp->v_mount->mnt_vnodecovered != NULLVP)) {
vnode_t coveredvp;
/* switch fvp to the covered vnode */
coveredvp = fvp->v_mount->mnt_vnodecovered;
if ((vnode_getwithref(coveredvp))) {
error = ENOENT;
goto out1;
}
/*
* Save the 'fvp' as it is needed for vn_authorize_renamex_with_paths()
* later.
*/
mnt_fvp = fvp;
fvp = coveredvp;
mntrename = TRUE;
}
/*
* Check for cross-device rename.
*/
if ((fvp->v_mount != tdvp->v_mount) ||
(tvp && (fvp->v_mount != tvp->v_mount))) {
error = EXDEV;
goto out1;
}
/*
* If source is the same as the destination (that is the
* same inode number) then there is nothing to do...
* EXCEPT if the underlying file system supports case
* insensitivity and is case preserving. In this case
* the file system needs to handle the special case of
* getting the same vnode as target (fvp) and source (tvp).
*
* Only file systems that support pathconf selectors _PC_CASE_SENSITIVE
* and _PC_CASE_PRESERVING can have this exception, and they need to
* handle the special case of getting the same vnode as target and
* source. NOTE: Then the target is unlocked going into vnop_rename,
* so not to cause locking problems. There is a single reference on tvp.
*
* NOTE - that fvp == tvp also occurs if they are hard linked and
* that correct behaviour then is just to return success without doing
* anything.
*
* XXX filesystem should take care of this itself, perhaps...
*/
if (fvp == tvp && fdvp == tdvp) {
if (fromnd->ni_cnd.cn_namelen == tond->ni_cnd.cn_namelen &&
!bcmp(fromnd->ni_cnd.cn_nameptr, tond->ni_cnd.cn_nameptr,
fromnd->ni_cnd.cn_namelen)) {
vn_authorize_skipped = TRUE;
goto out1;
}
}
if (holding_mntlock && fvp->v_mount != locked_mp) {
/*
* we're holding a reference and lock
* on locked_mp, but it no longer matches
* what we want to do... so drop our hold
*/
mount_unlock_renames(locked_mp);
mount_drop(locked_mp, 0);
holding_mntlock = 0;
}
if (tdvp != fdvp && fvp->v_type == VDIR) {
/*
* serialize renames that re-shape
* the tree... if holding_mntlock is
* set, then we're ready to go...
* otherwise we
* first need to drop the iocounts
* we picked up, second take the
* lock to serialize the access,
* then finally start the lookup
* process over with the lock held
*/
if (!holding_mntlock) {
/*
* need to grab a reference on
* the mount point before we
* drop all the iocounts... once
* the iocounts are gone, the mount
* could follow
*/
locked_mp = fvp->v_mount;
mount_ref(locked_mp, 0);
/*
* nameidone has to happen before we vnode_put(tvp)
* since it may need to release the fs_nodelock on the tvp
*/
nameidone(tond);
if (tvp) {
vnode_put(tvp);
}
vnode_put(tdvp);
/*
* nameidone has to happen before we vnode_put(fdvp)
* since it may need to release the fs_nodelock on the fvp
*/
nameidone(fromnd);
vnode_put(fvp);
vnode_put(fdvp);
if (mnt_fvp != NULLVP) {
vnode_put(mnt_fvp);
}
mount_lock_renames(locked_mp);
holding_mntlock = 1;
goto retry;
}
} else {
/*
* when we dropped the iocounts to take
* the lock, we allowed the identity of
* the various vnodes to change... if they did,
* we may no longer be dealing with a rename
* that reshapes the tree... once we're holding
* the iocounts, the vnodes can't change type
* so we're free to drop the lock at this point
* and continue on
*/
if (holding_mntlock) {
mount_unlock_renames(locked_mp);
mount_drop(locked_mp, 0);
holding_mntlock = 0;
}
}
if (!batched) {
error = vn_authorize_renamex_with_paths(fdvp, mntrename ? mnt_fvp : fvp,
&fromnd->ni_cnd, from_name, tdvp, tvp, &tond->ni_cnd, to_name, ctx,
flags, NULL);
if (error) {
if (error == ENOENT) {
if (retry_count < MAX_AUTHORIZE_ENOENT_RETRIES) {
/*
* We encountered a race where after doing the namei,
* tvp stops being valid. If so, simply re-drive the rename
* call from the top.
*/
do_retry = 1;
retry_count += 1;
}
}
goto out1;
}
}
/* Release the 'mnt_fvp' now that it is no longer needed. */
if (mnt_fvp != NULLVP) {
vnode_put(mnt_fvp);
mnt_fvp = NULLVP;
}
// save these off so we can later verify that fvp is the same
oname = fvp->v_name;
oparent = fvp->v_parent;
skipped_lookup:
#if CONFIG_FILE_LEASES
/* Lease break needed for source's parent dir? */
vnode_breakdirlease(fdvp, false, O_WRONLY);
/* Lease break needed for target's parent dir? */
vnode_breakdirlease(tdvp, false, O_WRONLY);
#endif
error = vn_rename(fdvp, &fvp, &fromnd->ni_cnd, fvap,
tdvp, &tvp, &tond->ni_cnd, tvap,
flags, ctx);
if (holding_mntlock) {
/*
* we can drop our serialization
* lock now
*/
mount_unlock_renames(locked_mp);
mount_drop(locked_mp, 0);
holding_mntlock = 0;
}
if (error) {
if (error == EDATALESS) {
/*
* If we've been here before, something has gone
* horribly wrong and we should just get out lest
* we spiral around the drain forever.
*/
if (flags & VFS_RENAME_DATALESS) {
error = EIO;
goto out1;
}
/*
* The object we're renaming is dataless (or has a
* dataless descendent) and requires materialization
* before the rename occurs. But we're holding the
* mount point's rename lock, so it's not safe to
* make the upcall.
*
* In this case, we release the lock (above), perform
* the materialization, and start the whole thing over.
*/
error = vfs_materialize_reparent(fvp, tdvp);
if (error == 0) {
/*
* The next time around we need to tell the
* file system that the materializtaion has
* been performed.
*/
flags |= VFS_RENAME_DATALESS;
do_retry = 1;
}
goto out1;
}
if (error == EKEEPLOOKING) {
if ((fromnd->ni_flag & NAMEI_CONTLOOKUP) == 0) {
if ((tond->ni_flag & NAMEI_CONTLOOKUP) == 0) {
panic("EKEEPLOOKING without NAMEI_CONTLOOKUP on either ndp?");
}
}
fromnd->ni_vp = fvp;
tond->ni_vp = tvp;
goto continue_lookup;
}
/*
* We may encounter a race in the VNOP where the destination didn't
* exist when we did the namei, but it does by the time we go and
* try to create the entry. In this case, we should re-drive this rename
* call from the top again. Currently, only HFS bubbles out ERECYCLE,
* but other filesystems susceptible to this race could return it, too.
*/
if (error == ERECYCLE) {
if (retry_count < MAX_RENAME_ERECYCLE_RETRIES) {
do_retry = 1;
retry_count += 1;
} else {
printf("rename retry limit due to ERECYCLE reached\n");
error = ENOENT;
}
}
/*
* For compound VNOPs, the authorization callback may return
* ENOENT in case of racing hardlink lookups hitting the name
* cache, redrive the lookup.
*/
if (batched && error == ENOENT) {
if (retry_count < MAX_AUTHORIZE_ENOENT_RETRIES) {
do_retry = 1;
retry_count += 1;
}
}
goto out1;
}
/* call out to allow 3rd party notification of rename.
* Ignore result of kauth_authorize_fileop call.
*/
kauth_authorize_fileop(vfs_context_ucred(ctx),
KAUTH_FILEOP_RENAME,
(uintptr_t)from_name, (uintptr_t)to_name);
if (flags & VFS_RENAME_SWAP) {
kauth_authorize_fileop(vfs_context_ucred(ctx),
KAUTH_FILEOP_RENAME,
(uintptr_t)to_name, (uintptr_t)from_name);
}
#if CONFIG_FSE
if (from_name != NULL && to_name != NULL) {
if (from_truncated || to_truncated) {
// set it here since only the from_finfo gets reported up to user space
from_finfo.mode |= FSE_TRUNCATED_PATH;
}
if (tvap && tvp) {
vnode_get_fse_info_from_vap(tvp, &to_finfo, tvap);
}
if (fvap) {
vnode_get_fse_info_from_vap(fvp, &from_finfo, fvap);
}
if (tvp) {
add_fsevent(FSE_RENAME, ctx,
FSE_ARG_STRING, from_len_no_firmlink, from_name_no_firmlink,
FSE_ARG_FINFO, &from_finfo,
FSE_ARG_STRING, to_len_no_firmlink, to_name_no_firmlink,
FSE_ARG_FINFO, &to_finfo,
FSE_ARG_DONE);
if (flags & VFS_RENAME_SWAP) {
/*
* Strictly speaking, swap is the equivalent of
* *three* renames. FSEvents clients should only take
* the events as a hint, so we only bother reporting
* two.
*/
add_fsevent(FSE_RENAME, ctx,
FSE_ARG_STRING, to_len_no_firmlink, to_name_no_firmlink,
FSE_ARG_FINFO, &to_finfo,
FSE_ARG_STRING, from_len_no_firmlink, from_name_no_firmlink,
FSE_ARG_FINFO, &from_finfo,
FSE_ARG_DONE);
}
} else {
add_fsevent(FSE_RENAME, ctx,
FSE_ARG_STRING, from_len_no_firmlink, from_name_no_firmlink,
FSE_ARG_FINFO, &from_finfo,
FSE_ARG_STRING, to_len_no_firmlink, to_name_no_firmlink,
FSE_ARG_DONE);
}
}
#endif /* CONFIG_FSE */
/*
* update filesystem's mount point data
*/
if (mntrename) {
char *cp, *pathend, *mpname;
char * tobuf;
struct mount *mp;
int maxlen;
size_t len = 0;
mp = fvp->v_mountedhere;
if (vfs_busy(mp, LK_NOWAIT)) {
error = EBUSY;
goto out1;
}
tobuf = zalloc(ZV_NAMEI);
if (UIO_SEG_IS_USER_SPACE(segflg)) {
error = copyinstr(to, tobuf, MAXPATHLEN, &len);
} else {
error = copystr((void *)to, tobuf, MAXPATHLEN, &len);
}
if (!error) {
/* find current mount point prefix */
pathend = &mp->mnt_vfsstat.f_mntonname[0];
for (cp = pathend; *cp != '\0'; ++cp) {
if (*cp == '/') {
pathend = cp + 1;
}
}
/* find last component of target name */
for (mpname = cp = tobuf; *cp != '\0'; ++cp) {
if (*cp == '/') {
mpname = cp + 1;
}
}
/* Update f_mntonname of sub mounts */
vfs_iterate(0, rename_submounts_callback, (void *)mp);
/* append name to prefix */
maxlen = MAXPATHLEN - (int)(pathend - mp->mnt_vfsstat.f_mntonname);
bzero(pathend, maxlen);
strlcpy(pathend, mpname, maxlen);
}
zfree(ZV_NAMEI, tobuf);
vfs_unbusy(mp);
vfs_event_signal(NULL, VQ_UPDATE, (intptr_t)NULL);
}
/*
* fix up name & parent pointers. note that we first
* check that fvp has the same name/parent pointers it
* had before the rename call... this is a 'weak' check
* at best...
*
* XXX oparent and oname may not be set in the compound vnop case
*/
if (batched || (oname == fvp->v_name && oparent == fvp->v_parent)) {
int update_flags;
update_flags = VNODE_UPDATE_NAME;
if (fdvp != tdvp) {
update_flags |= VNODE_UPDATE_PARENT;
}
vnode_update_identity(fvp, tdvp, tond->ni_cnd.cn_nameptr, tond->ni_cnd.cn_namelen, tond->ni_cnd.cn_hash, update_flags);
}
out1:
/*
* There are some cases (for e.g. 'fvp == tvp') when vn_authorize was
* skipped earlier as no actual rename was performed.
*/
if (vn_authorize_skipped && error == 0) {
error = vn_authorize_renamex_with_paths(fdvp, fvp,
&fromnd->ni_cnd, from_name, tdvp, tvp, &tond->ni_cnd, to_name, ctx,
flags, NULL);
if (error && error == ENOENT) {
if (retry_count < MAX_AUTHORIZE_ENOENT_RETRIES) {
do_retry = 1;
retry_count += 1;
}
}
}
if (to_name != NULL) {
RELEASE_PATH(to_name);
to_name = NULL;
}
if (to_name_no_firmlink != NULL) {
RELEASE_PATH(to_name_no_firmlink);
to_name_no_firmlink = NULL;
}
if (from_name != NULL) {
RELEASE_PATH(from_name);
from_name = NULL;
}
if (from_name_no_firmlink != NULL) {
RELEASE_PATH(from_name_no_firmlink);
from_name_no_firmlink = NULL;
}
if (holding_mntlock) {
mount_unlock_renames(locked_mp);
mount_drop(locked_mp, 0);
holding_mntlock = 0;
}
if (tdvp) {
/*
* nameidone has to happen before we vnode_put(tdvp)
* since it may need to release the fs_nodelock on the tdvp
*/
nameidone(tond);
if (tvp) {
vnode_put(tvp);
}
vnode_put(tdvp);
}
if (fdvp) {
/*
* nameidone has to happen before we vnode_put(fdvp)
* since it may need to release the fs_nodelock on the fdvp
*/
nameidone(fromnd);
if (fvp) {
vnode_put(fvp);
}
vnode_put(fdvp);
}
if (mnt_fvp != NULLVP) {
vnode_put(mnt_fvp);
}
/*
* If things changed after we did the namei, then we will re-drive
* this rename call from the top.
*/
if (do_retry) {
do_retry = 0;
goto retry;
}
kfree_type(typeof(*__rename_data), __rename_data);
return error;
}
int
rename(__unused proc_t p, struct rename_args *uap, __unused int32_t *retval)
{
return renameat_internal(vfs_context_current(), AT_FDCWD, uap->from,
AT_FDCWD, uap->to, UIO_USERSPACE, 0);
}
int
renameatx_np(__unused proc_t p, struct renameatx_np_args *uap, __unused int32_t *retval)
{
if (uap->flags & ~(RENAME_SECLUDE | RENAME_EXCL | RENAME_SWAP | RENAME_NOFOLLOW_ANY)) {
return EINVAL;
}
if ((uap->flags & (RENAME_EXCL | RENAME_SWAP)) == (RENAME_EXCL | RENAME_SWAP)) {
return EINVAL;
}
return renameat_internal(vfs_context_current(), uap->fromfd, uap->from,
uap->tofd, uap->to, UIO_USERSPACE, uap->flags);
}
int
renameat(__unused proc_t p, struct renameat_args *uap, __unused int32_t *retval)
{
return renameat_internal(vfs_context_current(), uap->fromfd, uap->from,
uap->tofd, uap->to, UIO_USERSPACE, 0);
}
/*
* Make a directory file.
*
* Returns: 0 Success
* EEXIST
* namei:???
* vnode_authorize:???
* vn_create:???
*/
/* ARGSUSED */
static int
mkdir1at(vfs_context_t ctx, user_addr_t path, struct vnode_attr *vap, int fd,
enum uio_seg segflg)
{
vnode_t vp, dvp;
int error;
int update_flags = 0;
int batched;
struct nameidata nd;
AUDIT_ARG(mode, vap->va_mode);
NDINIT(&nd, CREATE, OP_MKDIR, LOCKPARENT | AUDITVNPATH1, segflg,
path, ctx);
nd.ni_cnd.cn_flags |= WILLBEDIR;
nd.ni_flag = NAMEI_COMPOUNDMKDIR;
continue_lookup:
error = nameiat(&nd, fd);
if (error) {
return error;
}
dvp = nd.ni_dvp;
vp = nd.ni_vp;
if (vp != NULL) {
error = EEXIST;
goto out;
}
batched = vnode_compound_mkdir_available(dvp);
VATTR_SET(vap, va_type, VDIR);
/*
* XXX
* Don't authorize in VFS for compound VNOP.... mkdir -p today assumes that it will
* only get EXISTS or EISDIR for existing path components, and not that it could see
* EACCESS/EPERM--so if we authorize for mkdir on "/" for "mkdir -p /tmp/foo/bar/baz"
* it will fail in a spurious manner. Need to figure out if this is valid behavior.
*/
if ((error = vn_authorize_mkdir(dvp, &nd.ni_cnd, vap, ctx, NULL)) != 0) {
if (error == EACCES || error == EPERM) {
int error2;
nameidone(&nd);
vnode_put(dvp);
dvp = NULLVP;
/*
* Try a lookup without "NAMEI_COMPOUNDVNOP" to make sure we return EEXIST
* rather than EACCESS if the target exists.
*/
NDINIT(&nd, LOOKUP, OP_MKDIR, AUDITVNPATH1, segflg,
path, ctx);
error2 = nameiat(&nd, fd);
if (error2) {
goto out;
} else {
vp = nd.ni_vp;
error = EEXIST;
goto out;
}
}
goto out;
}
#if CONFIG_FILE_LEASES
vnode_breakdirlease(dvp, false, O_WRONLY);
#endif
/*
* make the directory
*/
if ((error = vn_create(dvp, &vp, &nd, vap, 0, 0, NULL, ctx)) != 0) {
if (error == EKEEPLOOKING) {
nd.ni_vp = vp;
goto continue_lookup;
}
goto out;
}
// Make sure the name & parent pointers are hooked up
if (vp->v_name == NULL) {
update_flags |= VNODE_UPDATE_NAME;
}
if (vp->v_parent == NULLVP) {
update_flags |= VNODE_UPDATE_PARENT;
}
if (update_flags) {
vnode_update_identity(vp, dvp, nd.ni_cnd.cn_nameptr, nd.ni_cnd.cn_namelen, nd.ni_cnd.cn_hash, update_flags);
}
#if CONFIG_FSE
add_fsevent(FSE_CREATE_DIR, ctx, FSE_ARG_VNODE, vp, FSE_ARG_DONE);
#endif
out:
/*
* nameidone has to happen before we vnode_put(dvp)
* since it may need to release the fs_nodelock on the dvp
*/
nameidone(&nd);
if (vp) {
vnode_put(vp);
}
if (dvp) {
vnode_put(dvp);
}
return error;
}
/*
* mkdir_extended: Create a directory; with extended security (ACL).
*
* Parameters: p Process requesting to create the directory
* uap User argument descriptor (see below)
* retval (ignored)
*
* Indirect: uap->path Path of directory to create
* uap->mode Access permissions to set
* uap->xsecurity ACL to set
*
* Returns: 0 Success
* !0 Not success
*
*/
int
mkdir_extended(proc_t p, struct mkdir_extended_args *uap, __unused int32_t *retval)
{
int ciferror;
kauth_filesec_t xsecdst;
struct vnode_attr va;
AUDIT_ARG(owner, uap->uid, uap->gid);
xsecdst = NULL;
if ((uap->xsecurity != USER_ADDR_NULL) &&
((ciferror = kauth_copyinfilesec(uap->xsecurity, &xsecdst)) != 0)) {
return ciferror;
}
VATTR_INIT(&va);
VATTR_SET(&va, va_mode, (uap->mode & ACCESSPERMS) & ~p->p_fd.fd_cmask);
if (xsecdst != NULL) {
VATTR_SET(&va, va_acl, &xsecdst->fsec_acl);
va.va_vaflags |= VA_FILESEC_ACL;
}
ciferror = mkdir1at(vfs_context_current(), uap->path, &va, AT_FDCWD,
UIO_USERSPACE);
if (xsecdst != NULL) {
kauth_filesec_free(xsecdst);
}
return ciferror;
}
int
mkdir(proc_t p, struct mkdir_args *uap, __unused int32_t *retval)
{
struct vnode_attr va;
VATTR_INIT(&va);
VATTR_SET(&va, va_mode, (uap->mode & ACCESSPERMS) & ~p->p_fd.fd_cmask);
return mkdir1at(vfs_context_current(), uap->path, &va, AT_FDCWD,
UIO_USERSPACE);
}
int
mkdirat(proc_t p, struct mkdirat_args *uap, __unused int32_t *retval)
{
struct vnode_attr va;
VATTR_INIT(&va);
VATTR_SET(&va, va_mode, (uap->mode & ACCESSPERMS) & ~p->p_fd.fd_cmask);
return mkdir1at(vfs_context_current(), uap->path, &va, uap->fd,
UIO_USERSPACE);
}
static int
rmdirat_internal(vfs_context_t ctx, int fd, user_addr_t dirpath,
enum uio_seg segflg, int unlink_flags)
{
struct {
struct nameidata nd;
#if CONFIG_FSE
struct vnode_attr va;
#endif /* CONFIG_FSE */
} *__rmdir_data;
vnode_t vp, dvp;
int error;
struct nameidata *ndp;
char *path = NULL;
char *no_firmlink_path = NULL;
int len_path = 0;
int len_no_firmlink_path = 0;
int has_listeners = 0;
int need_event = 0;
int truncated_path = 0;
int truncated_no_firmlink_path = 0;
struct vnode_attr *vap = NULL;
int restart_count = 0;
int batched;
int restart_flag;
int nofollow_any = 0;
__rmdir_data = kalloc_type(typeof(*__rmdir_data), Z_WAITOK);
ndp = &__rmdir_data->nd;
if (unlink_flags & VNODE_REMOVE_NOFOLLOW_ANY) {
nofollow_any = NAMEI_NOFOLLOW_ANY;
unlink_flags &= ~VNODE_REMOVE_NOFOLLOW_ANY;
}
/*
* This loop exists to restart rmdir in the unlikely case that two
* processes are simultaneously trying to remove the same directory
* containing orphaned appleDouble files.
*/
do {
NDINIT(ndp, DELETE, OP_RMDIR, LOCKPARENT | AUDITVNPATH1,
segflg, dirpath, ctx);
ndp->ni_flag = NAMEI_COMPOUNDRMDIR | nofollow_any;
continue_lookup:
restart_flag = 0;
vap = NULL;
error = nameiat(ndp, fd);
if (error) {
goto err_out;
}
dvp = ndp->ni_dvp;
vp = ndp->ni_vp;
if (vp) {
batched = vnode_compound_rmdir_available(vp);
if (vp->v_flag & VROOT) {
/*
* The root of a mounted filesystem cannot be deleted.
*/
error = EBUSY;
goto out;
}
#if DEVELOPMENT || DEBUG
/*
* XXX VSWAP: Check for entitlements or special flag here
* so we can restrict access appropriately.
*/
#else /* DEVELOPMENT || DEBUG */
if (vnode_isswap(vp) && (ctx != vfs_context_kernel())) {
error = EPERM;
goto out;
}
#endif /* DEVELOPMENT || DEBUG */
/*
* Removed a check here; we used to abort if vp's vid
* was not the same as what we'd seen the last time around.
* I do not think that check was valid, because if we retry
* and all dirents are gone, the directory could legitimately
* be recycled but still be present in a situation where we would
* have had permission to delete. Therefore, we won't make
* an effort to preserve that check now that we may not have a
* vp here.
*/
if (!batched) {
error = vn_authorize_rmdir(dvp, vp, &ndp->ni_cnd, ctx, NULL);
if (error) {
if (error == ENOENT) {
if (restart_count < MAX_AUTHORIZE_ENOENT_RETRIES) {
restart_flag = 1;
restart_count += 1;
}
}
goto out;
}
}
} else {
batched = 1;
if (!vnode_compound_rmdir_available(dvp)) {
panic("No error, but no compound rmdir?");
}
}
#if CONFIG_FSE
fse_info finfo = {0};
need_event = need_fsevent(FSE_DELETE, dvp);
if (need_event) {
if (!batched) {
get_fse_info(vp, &finfo, ctx);
} else {
error = vfs_get_notify_attributes(&__rmdir_data->va);
if (error) {
goto out;
}
vap = &__rmdir_data->va;
}
}
#endif
has_listeners = kauth_authorize_fileop_has_listeners();
if (need_event || has_listeners) {
if (path == NULL) {
GET_PATH(path);
}
len_path = safe_getpath(dvp, ndp->ni_cnd.cn_nameptr, path, MAXPATHLEN, &truncated_path);
if (no_firmlink_path == NULL) {
GET_PATH(no_firmlink_path);
}
len_no_firmlink_path = safe_getpath_no_firmlink(dvp, ndp->ni_cnd.cn_nameptr, no_firmlink_path, MAXPATHLEN, &truncated_no_firmlink_path);
#if CONFIG_FSE
if (truncated_no_firmlink_path) {
finfo.mode |= FSE_TRUNCATED_PATH;
}
#endif
}
#if CONFIG_FILE_LEASES
vnode_breakdirlease(dvp, false, O_WRONLY);
#endif
error = vn_rmdir(dvp, &vp, ndp, vap, ctx);
ndp->ni_vp = vp;
if (vp == NULLVP) {
/* Couldn't find a vnode */
goto out;
}
if (error == EKEEPLOOKING) {
goto continue_lookup;
} else if (batched && error == ENOENT) {
if (restart_count < MAX_AUTHORIZE_ENOENT_RETRIES) {
/*
* For compound VNOPs, the authorization callback
* may return ENOENT in case of racing hard link lookups
* redrive the lookup.
*/
restart_flag = 1;
restart_count += 1;
goto out;
}
}
/*
* XXX There's no provision for passing flags
* to VNOP_RMDIR(). So, if vn_rmdir() fails
* because it's not empty, then we try again
* with VNOP_REMOVE(), passing in a special
* flag that clever file systems will know
* how to handle.
*/
if (error == ENOTEMPTY &&
(unlink_flags & VNODE_REMOVE_DATALESS_DIR) != 0) {
/*
* Only do this if the directory is actually
* marked as DATALESS.
*/
struct vnode_attr *lvap =
kalloc_type(struct vnode_attr, Z_WAITOK);
VATTR_INIT(lvap);
VATTR_WANTED(lvap, va_flags);
if (vnode_getattr(vp, lvap, ctx) == 0 &&
VATTR_IS_SUPPORTED(lvap, va_flags) &&
(lvap->va_flags & SF_DATALESS) != 0) {
/*
* If this fails, we want to keep the original
* error.
*/
if (vn_remove(dvp, &vp, ndp,
VNODE_REMOVE_DATALESS_DIR, vap, ctx) == 0) {
error = 0;
}
}
kfree_type(struct vnode_attr, lvap);
}
#if CONFIG_APPLEDOUBLE
/*
* Special case to remove orphaned AppleDouble
* files. I don't like putting this in the kernel,
* but carbon does not like putting this in carbon either,
* so here we are.
*/
if (error == ENOTEMPTY) {
int ad_error = rmdir_remove_orphaned_appleDouble(vp, ctx, &restart_flag);
if (ad_error == EBUSY) {
error = ad_error;
goto out;
}
/*
* Assuming everything went well, we will try the RMDIR again
*/
if (!ad_error) {
error = vn_rmdir(dvp, &vp, ndp, vap, ctx);
}
}
#endif /* CONFIG_APPLEDOUBLE */
/*
* Call out to allow 3rd party notification of delete.
* Ignore result of kauth_authorize_fileop call.
*/
if (!error) {
if (has_listeners) {
kauth_authorize_fileop(vfs_context_ucred(ctx),
KAUTH_FILEOP_DELETE,
(uintptr_t)vp,
(uintptr_t)path);
}
if (vp->v_flag & VISHARDLINK) {
// see the comment in unlink1() about why we update
// the parent of a hard link when it is removed
vnode_update_identity(vp, NULL, NULL, 0, 0, VNODE_UPDATE_PARENT);
}
#if CONFIG_FSE
if (need_event) {
if (vap) {
vnode_get_fse_info_from_vap(vp, &finfo, vap);
}
add_fsevent(FSE_DELETE, ctx,
FSE_ARG_STRING, len_no_firmlink_path, no_firmlink_path,
FSE_ARG_FINFO, &finfo,
FSE_ARG_DONE);
}
#endif
#if CONFIG_MACF
mac_vnode_notify_unlink(ctx, dvp, vp, &ndp->ni_cnd);
#endif
}
out:
if (path != NULL) {
RELEASE_PATH(path);
path = NULL;
}
if (no_firmlink_path != NULL) {
RELEASE_PATH(no_firmlink_path);
no_firmlink_path = NULL;
}
/*
* nameidone has to happen before we vnode_put(dvp)
* since it may need to release the fs_nodelock on the dvp
*/
nameidone(ndp);
vnode_put(dvp);
if (vp) {
vnode_put(vp);
}
if (restart_flag == 0) {
wakeup_one((caddr_t)vp);
goto err_out;
}
tsleep(vp, PVFS, "rm AD", 1);
} while (restart_flag != 0);
err_out:
kfree_type(typeof(*__rmdir_data), __rmdir_data);
return error;
}
/*
* Remove a directory file.
*/
/* ARGSUSED */
int
rmdir(__unused proc_t p, struct rmdir_args *uap, __unused int32_t *retval)
{
return rmdirat_internal(vfs_context_current(), AT_FDCWD,
CAST_USER_ADDR_T(uap->path), UIO_USERSPACE, 0);
}
/* Get direntry length padded to 8 byte alignment */
#define DIRENT64_LEN(namlen) \
((sizeof(struct direntry) + (namlen) - (MAXPATHLEN-1) + 7) & ~7)
/* Get dirent length padded to 4 byte alignment */
#define DIRENT_LEN(namelen) \
((sizeof(struct dirent) + (namelen + 1) - (__DARWIN_MAXNAMLEN + 1) + 3) & ~3)
/* Get the end of this dirent */
#define DIRENT_END(dep) \
(((char *)(dep)) + (dep)->d_reclen - 1)
errno_t
vnode_readdir64(struct vnode *vp, struct uio *uio, int flags, int *eofflag,
int *numdirent, vfs_context_t ctxp)
{
/* Check if fs natively supports VNODE_READDIR_EXTENDED */
if ((vp->v_mount->mnt_vtable->vfc_vfsflags & VFC_VFSREADDIR_EXTENDED) &&
((vp->v_mount->mnt_kern_flag & MNTK_DENY_READDIREXT) == 0)) {
return VNOP_READDIR(vp, uio, flags, eofflag, numdirent, ctxp);
} else {
size_t bufsize;
void * bufptr;
uio_t auio;
struct direntry *entry64;
struct dirent *dep;
size_t bytesread;
int error;
/*
* We're here because the underlying file system does not
* support direnties or we mounted denying support so we must
* fall back to dirents and convert them to direntries.
*
* Our kernel buffer needs to be smaller since re-packing will
* expand each dirent. The worse case (when the name length
* is 3 or less) corresponds to a struct direntry size of 32
* bytes (8-byte aligned) and a struct dirent size of 12 bytes
* (4-byte aligned). So having a buffer that is 3/8 the size
* will prevent us from reading more than we can pack.
*
* Since this buffer is wired memory, we will limit the
* buffer size to a maximum of 32K. We would really like to
* use 32K in the MIN(), but we use magic number 87371 to
* prevent uio_resid() * 3 / 8 from overflowing.
*/
bufsize = 3 * MIN((user_size_t)uio_resid(uio), 87371u) / 8;
bufptr = kalloc_data(bufsize, Z_WAITOK);
if (bufptr == NULL) {
return ENOMEM;
}
auio = uio_create(1, 0, UIO_SYSSPACE, UIO_READ);
uio_addiov(auio, (uintptr_t)bufptr, bufsize);
auio->uio_offset = uio->uio_offset;
error = VNOP_READDIR(vp, auio, 0, eofflag, numdirent, ctxp);
dep = (struct dirent *)bufptr;
bytesread = bufsize - uio_resid(auio);
entry64 = kalloc_type(struct direntry, Z_WAITOK);
/*
* Convert all the entries and copy them out to user's buffer.
*/
while (error == 0 && (char *)dep < ((char *)bufptr + bytesread)) {
/* First check that the dirent struct up to d_name is within the buffer */
if ((char*)dep + offsetof(struct dirent, d_name) > ((char *)bufptr + bytesread) ||
/* Check that the length of the entire dirent is within the buffer */
DIRENT_END(dep) > ((char *)bufptr + bytesread) ||
/* Check that the actual length including the name doesn't exceed d_reclen */
DIRENT_LEN(dep->d_namlen) > dep->d_reclen) {
printf("%s: %s: Bad dirent recived from directory %s\n", __func__,
vp->v_mount->mnt_vfsstat.f_mntonname,
vp->v_name ? vp->v_name : "<unknown>");
error = EIO;
break;
}
size_t enbufsize = DIRENT64_LEN(dep->d_namlen);
bzero(entry64, enbufsize);
/* Convert a dirent to a dirent64. */
entry64->d_ino = dep->d_ino;
entry64->d_seekoff = 0;
entry64->d_reclen = (uint16_t)enbufsize;
entry64->d_namlen = dep->d_namlen;
entry64->d_type = dep->d_type;
bcopy(dep->d_name, entry64->d_name, dep->d_namlen + 1);
/* Move to next entry. */
dep = (struct dirent *)((char *)dep + dep->d_reclen);
/* Copy entry64 to user's buffer. */
error = uiomove((caddr_t)entry64, entry64->d_reclen, uio);
}
/* Update the real offset using the offset we got from VNOP_READDIR. */
if (error == 0) {
uio->uio_offset = auio->uio_offset;
}
uio_free(auio);
kfree_data(bufptr, bufsize);
kfree_type(struct direntry, entry64);
return error;
}
}
#define GETDIRENTRIES_MAXBUFSIZE (128 * 1024 * 1024U)
/*
* Read a block of directory entries in a file system independent format.
*/
static int
getdirentries_common(int fd, user_addr_t bufp, user_size_t bufsize, ssize_t *bytesread,
off_t *offset, int *eofflag, int flags)
{
vnode_t vp;
struct vfs_context context = *vfs_context_current(); /* local copy */
struct fileproc *fp;
uio_t auio;
int spacetype = proc_is64bit(vfs_context_proc(&context)) ? UIO_USERSPACE64 : UIO_USERSPACE32;
off_t loff;
int error, numdirent;
UIO_STACKBUF(uio_buf, 1);
get_from_fd:
error = fp_getfvp(vfs_context_proc(&context), fd, &fp, &vp);
if (error) {
return error;
}
vn_offset_lock(fp->fp_glob);
if (((vnode_t)fp_get_data(fp)) != vp) {
vn_offset_unlock(fp->fp_glob);
file_drop(fd);
goto get_from_fd;
}
if ((fp->fp_glob->fg_flag & FREAD) == 0) {
AUDIT_ARG(vnpath_withref, vp, ARG_VNODE1);
error = EBADF;
goto out;
}
if (bufsize > GETDIRENTRIES_MAXBUFSIZE) {
bufsize = GETDIRENTRIES_MAXBUFSIZE;
}
#if CONFIG_MACF
error = mac_file_check_change_offset(vfs_context_ucred(&context), fp->fp_glob);
if (error) {
goto out;
}
#endif
if ((error = vnode_getwithref(vp))) {
goto out;
}
AUDIT_ARG(vnpath, vp, ARG_VNODE1);
#if CONFIG_UNION_MOUNTS
unionread:
#endif /* CONFIG_UNION_MOUNTS */
if (vp->v_type != VDIR) {
(void)vnode_put(vp);
error = EINVAL;
goto out;
}
#if CONFIG_MACF
error = mac_vnode_check_readdir(&context, vp);
if (error != 0) {
(void)vnode_put(vp);
goto out;
}
#endif /* MAC */
loff = fp->fp_glob->fg_offset;
auio = uio_createwithbuffer(1, loff, spacetype, UIO_READ, &uio_buf[0], sizeof(uio_buf));
uio_addiov(auio, bufp, bufsize);
if (flags & VNODE_READDIR_EXTENDED) {
error = vnode_readdir64(vp, auio, flags, eofflag, &numdirent, &context);
fp->fp_glob->fg_offset = uio_offset(auio);
} else {
error = VNOP_READDIR(vp, auio, 0, eofflag, &numdirent, &context);
fp->fp_glob->fg_offset = uio_offset(auio);
}
if (error) {
(void)vnode_put(vp);
goto out;
}
#if CONFIG_UNION_MOUNTS
if ((user_ssize_t)bufsize == uio_resid(auio) &&
(vp->v_mount->mnt_flag & MNT_UNION)) {
vnode_t uvp;
if (lookup_traverse_union(vp, &uvp, &context) == 0) {
if (vnode_ref(uvp) == 0) {
fp_set_data(fp, uvp);
fp->fp_glob->fg_offset = 0;
vnode_rele(vp);
vnode_put(vp);
vp = uvp;
goto unionread;
} else {
/* could not get a ref, can't replace in fd */
vnode_put(uvp);
}
}
}
#endif /* CONFIG_UNION_MOUNTS */
vnode_put(vp);
if (offset) {
*offset = loff;
}
*bytesread = bufsize - uio_resid(auio);
out:
vn_offset_unlock(fp->fp_glob);
file_drop(fd);
return error;
}
int
getdirentries(__unused struct proc *p, struct getdirentries_args *uap, int32_t *retval)
{
off_t offset;
ssize_t bytesread;
int error, eofflag;
AUDIT_ARG(fd, uap->fd);
error = getdirentries_common(uap->fd, uap->buf, uap->count,
&bytesread, &offset, &eofflag, 0);
if (error == 0) {
if (proc_is64bit(p)) {
user64_long_t base = (user64_long_t)offset;
error = copyout((caddr_t)&base, uap->basep, sizeof(user64_long_t));
} else {
user32_long_t base = (user32_long_t)offset;
error = copyout((caddr_t)&base, uap->basep, sizeof(user32_long_t));
}
*retval = (int)bytesread;
}
return error;
}
int
getdirentries64(__unused struct proc *p, struct getdirentries64_args *uap, user_ssize_t *retval)
{
off_t offset;
ssize_t bytesread;
int error, eofflag;
user_size_t bufsize;
AUDIT_ARG(fd, uap->fd);
/*
* If the buffer is at least GETDIRENTRIES64_EXTENDED_BUFSIZE large,
* then the kernel carves out the last 4 bytes to return extended
* information to userspace (namely whether we reached EOF with this call).
*/
if (uap->bufsize >= GETDIRENTRIES64_EXTENDED_BUFSIZE) {
bufsize = uap->bufsize - sizeof(getdirentries64_flags_t);
} else {
bufsize = uap->bufsize;
}
error = getdirentries_common(uap->fd, uap->buf, bufsize,
&bytesread, &offset, &eofflag, VNODE_READDIR_EXTENDED);
if (error == 0) {
*retval = bytesread;
error = copyout((caddr_t)&offset, uap->position, sizeof(off_t));
if (error == 0 && uap->bufsize >= GETDIRENTRIES64_EXTENDED_BUFSIZE) {
getdirentries64_flags_t flags = 0;
if (eofflag) {
flags |= GETDIRENTRIES64_EOF;
}
error = copyout(&flags, (user_addr_t)uap->buf + bufsize,
sizeof(flags));
}
}
return error;
}
/*
* Set the mode mask for creation of filesystem nodes.
* XXX implement xsecurity
*/
#define UMASK_NOXSECURITY (void *)1 /* leave existing xsecurity alone */
static int
umask1(proc_t p, int newmask, __unused kauth_filesec_t fsec, int32_t *retval)
{
AUDIT_ARG(mask, newmask);
proc_fdlock(p);
*retval = p->p_fd.fd_cmask;
p->p_fd.fd_cmask = newmask & ALLPERMS;
proc_fdunlock(p);
return 0;
}
/*
* umask_extended: Set the mode mask for creation of filesystem nodes; with extended security (ACL).
*
* Parameters: p Process requesting to set the umask
* uap User argument descriptor (see below)
* retval umask of the process (parameter p)
*
* Indirect: uap->newmask umask to set
* uap->xsecurity ACL to set
*
* Returns: 0 Success
* !0 Not success
*
*/
int
umask_extended(proc_t p, struct umask_extended_args *uap, int32_t *retval)
{
return umask1(p, uap->newmask, KAUTH_FILESEC_NONE, retval);
}
int
umask(proc_t p, struct umask_args *uap, int32_t *retval)
{
return umask1(p, uap->newmask, UMASK_NOXSECURITY, retval);
}
#define REVOKE_MOUNTED_DEVICE_ENTITLEMENT \
"com.apple.private.vfs.revoke-mounted-device"
/*
* Void all references to file by ripping underlying filesystem
* away from vnode.
*/
/* ARGSUSED */
int
revoke(proc_t p, struct revoke_args *uap, __unused int32_t *retval)
{
vnode_t vp;
struct vnode_attr va;
vfs_context_t ctx = vfs_context_current();
int error;
struct nameidata nd;
NDINIT(&nd, LOOKUP, OP_REVOKE, FOLLOW | AUDITVNPATH1, UIO_USERSPACE,
uap->path, ctx);
error = namei(&nd);
if (error) {
return error;
}
vp = nd.ni_vp;
nameidone(&nd);
if (!(vnode_ischr(vp) || vnode_isblk(vp))) {
error = ENOTSUP;
goto out;
}
if (vnode_isblk(vp) && vnode_ismountedon(vp)) {
error = EBUSY;
goto out;
}
#if CONFIG_MACF
error = mac_vnode_check_revoke(ctx, vp);
if (error) {
goto out;
}
#endif
VATTR_INIT(&va);
VATTR_WANTED(&va, va_uid);
if ((error = vnode_getattr(vp, &va, ctx))) {
goto out;
}
if (kauth_cred_getuid(vfs_context_ucred(ctx)) != va.va_uid &&
(error = suser(vfs_context_ucred(ctx), &p->p_acflag))) {
goto out;
}
if (vp->v_usecount > 0 || (vnode_isaliased(vp))) {
VNOP_REVOKE(vp, REVOKEALL, ctx);
}
out:
vnode_put(vp);
return error;
}
/*
* HFS/HFS PlUS SPECIFIC SYSTEM CALLS
* The following system calls are designed to support features
* which are specific to the HFS & HFS Plus volume formats
*/
/*
* Obtain attribute information on objects in a directory while enumerating
* the directory.
*/
/* ARGSUSED */
int
getdirentriesattr(proc_t p, struct getdirentriesattr_args *uap, int32_t *retval)
{
vnode_t vp;
struct fileproc *fp;
uio_t auio = NULL;
int spacetype = proc_is64bit(p) ? UIO_USERSPACE64 : UIO_USERSPACE32;
uint32_t count = 0, savecount = 0;
uint32_t newstate = 0;
int error, eofflag = 0;
off_t loff = 0;
struct attrlist attributelist;
vfs_context_t ctx = vfs_context_current();
int fd = uap->fd;
UIO_STACKBUF(uio_buf, 1);
kauth_action_t action;
AUDIT_ARG(fd, fd);
/* Get the attributes into kernel space */
if ((error = copyin(uap->alist, (caddr_t)&attributelist, sizeof(attributelist)))) {
return error;
}
if ((error = copyin(uap->count, (caddr_t)&count, sizeof(count)))) {
return error;
}
savecount = count;
get_from_fd:
if ((error = fp_getfvp(p, fd, &fp, &vp))) {
return error;
}
vn_offset_lock(fp->fp_glob);
if (((vnode_t)fp_get_data(fp)) != vp) {
vn_offset_unlock(fp->fp_glob);
file_drop(fd);
goto get_from_fd;
}
if ((fp->fp_glob->fg_flag & FREAD) == 0) {
AUDIT_ARG(vnpath_withref, vp, ARG_VNODE1);
error = EBADF;
goto out;
}
#if CONFIG_MACF
error = mac_file_check_change_offset(vfs_context_ucred(ctx),
fp->fp_glob);
if (error) {
goto out;
}
#endif
if ((error = vnode_getwithref(vp))) {
goto out;
}
AUDIT_ARG(vnpath, vp, ARG_VNODE1);
#if CONFIG_UNION_MOUNTS
unionread:
#endif /* CONFIG_UNION_MOUNTS */
if (vp->v_type != VDIR) {
(void)vnode_put(vp);
error = EINVAL;
goto out;
}
#if CONFIG_MACF
error = mac_vnode_check_readdir(ctx, vp);
if (error != 0) {
(void)vnode_put(vp);
goto out;
}
#endif /* MAC */
/* set up the uio structure which will contain the users return buffer */
loff = fp->fp_glob->fg_offset;
auio = uio_createwithbuffer(1, loff, spacetype, UIO_READ, &uio_buf[0], sizeof(uio_buf));
uio_addiov(auio, uap->buffer, uap->buffersize);
/*
* If the only item requested is file names, we can let that past with
* just LIST_DIRECTORY. If they want any other attributes, that means
* they need SEARCH as well.
*/
action = KAUTH_VNODE_LIST_DIRECTORY;
if ((attributelist.commonattr & ~ATTR_CMN_NAME) ||
attributelist.fileattr || attributelist.dirattr) {
action |= KAUTH_VNODE_SEARCH;
}
if ((error = vnode_authorize(vp, NULL, action, ctx)) == 0) {
/* Believe it or not, uap->options only has 32-bits of valid
* info, so truncate before extending again */
error = VNOP_READDIRATTR(vp, &attributelist, auio, count,
(uint32_t)uap->options, &newstate, &eofflag, &count, ctx);
}
if (error) {
(void) vnode_put(vp);
goto out;
}
#if CONFIG_UNION_MOUNTS
/*
* If we've got the last entry of a directory in a union mount
* then reset the eofflag and pretend there's still more to come.
* The next call will again set eofflag and the buffer will be empty,
* so traverse to the underlying directory and do the directory
* read there.
*/
if (eofflag && vp->v_mount->mnt_flag & MNT_UNION) {
if (uio_resid(auio) < (user_ssize_t) uap->buffersize) { // Got some entries
eofflag = 0;
} else { // Empty buffer
vnode_t uvp;
if (lookup_traverse_union(vp, &uvp, ctx) == 0) {
if (vnode_ref_ext(uvp, fp->fp_glob->fg_flag & O_EVTONLY, 0) == 0) {
fp_set_data(fp, uvp);
fp->fp_glob->fg_offset = 0; // reset index for new dir
count = savecount;
vnode_rele_internal(vp, fp->fp_glob->fg_flag & O_EVTONLY, 0, 0);
vnode_put(vp);
vp = uvp;
goto unionread;
} else {
/* could not get a ref, can't replace in fd */
vnode_put(uvp);
}
}
}
}
#endif /* CONFIG_UNION_MOUNTS */
(void)vnode_put(vp);
if (error) {
goto out;
}
fp->fp_glob->fg_offset = uio_offset(auio); /* should be multiple of dirent, not variable */
if ((error = copyout((caddr_t) &count, uap->count, sizeof(count)))) {
goto out;
}
if ((error = copyout((caddr_t) &newstate, uap->newstate, sizeof(newstate)))) {
goto out;
}
if ((error = copyout((caddr_t) &loff, uap->basep, sizeof(loff)))) {
goto out;
}
*retval = eofflag; /* similar to getdirentries */
error = 0;
out:
vn_offset_unlock(fp->fp_glob);
file_drop(fd);
return error; /* return error earlier, an retval of 0 or 1 now */
} /* end of getdirentriesattr system call */
/*
* Exchange data between two files
*/
/* ARGSUSED */
int
exchangedata(__unused proc_t p, struct exchangedata_args *uap, __unused int32_t *retval)
{
struct nameidata fnd, snd;
vfs_context_t ctx = vfs_context_current();
vnode_t fvp;
vnode_t svp;
int error;
u_int32_t nameiflags;
char *fpath = NULL;
char *spath = NULL;
int flen = 0, slen = 0;
int from_truncated = 0, to_truncated = 0;
#if CONFIG_FSE
fse_info f_finfo, s_finfo;
#endif
nameiflags = 0;
if ((uap->options & FSOPT_NOFOLLOW) == 0) {
nameiflags |= FOLLOW;
}
NDINIT(&fnd, LOOKUP, OP_EXCHANGEDATA, nameiflags | AUDITVNPATH1,
UIO_USERSPACE, uap->path1, ctx);
error = namei(&fnd);
if (error) {
goto out2;
}
nameidone(&fnd);
fvp = fnd.ni_vp;
NDINIT(&snd, LOOKUP, OP_EXCHANGEDATA, CN_NBMOUNTLOOK | nameiflags | AUDITVNPATH2,
UIO_USERSPACE, uap->path2, ctx);
error = namei(&snd);
if (error) {
vnode_put(fvp);
goto out2;
}
nameidone(&snd);
svp = snd.ni_vp;
/*
* if the files are the same, return an inval error
*/
if (svp == fvp) {
error = EINVAL;
goto out;
}
/*
* if the files are on different volumes, return an error
*/
if (svp->v_mount != fvp->v_mount) {
error = EXDEV;
goto out;
}
/* If they're not files, return an error */
if ((vnode_isreg(fvp) == 0) || (vnode_isreg(svp) == 0)) {
error = EINVAL;
goto out;
}
#if CONFIG_MACF
error = mac_vnode_check_exchangedata(ctx,
fvp, svp);
if (error) {
goto out;
}
#endif
if (((error = vnode_authorize(fvp, NULL, KAUTH_VNODE_READ_DATA | KAUTH_VNODE_WRITE_DATA, ctx)) != 0) ||
((error = vnode_authorize(svp, NULL, KAUTH_VNODE_READ_DATA | KAUTH_VNODE_WRITE_DATA, ctx)) != 0)) {
goto out;
}
if (
#if CONFIG_FSE
need_fsevent(FSE_EXCHANGE, fvp) ||
#endif
kauth_authorize_fileop_has_listeners()) {
GET_PATH(fpath);
GET_PATH(spath);
flen = safe_getpath(fvp, NULL, fpath, MAXPATHLEN, &from_truncated);
slen = safe_getpath(svp, NULL, spath, MAXPATHLEN, &to_truncated);
#if CONFIG_FSE
get_fse_info(fvp, &f_finfo, ctx);
get_fse_info(svp, &s_finfo, ctx);
if (from_truncated || to_truncated) {
// set it here since only the f_finfo gets reported up to user space
f_finfo.mode |= FSE_TRUNCATED_PATH;
}
#endif
}
/* Ok, make the call */
error = VNOP_EXCHANGE(fvp, svp, 0, ctx);
if (error == 0) {
const char *tmpname;
if (fpath != NULL && spath != NULL) {
/* call out to allow 3rd party notification of exchangedata.
* Ignore result of kauth_authorize_fileop call.
*/
kauth_authorize_fileop(vfs_context_ucred(ctx), KAUTH_FILEOP_EXCHANGE,
(uintptr_t)fpath, (uintptr_t)spath);
}
name_cache_lock();
tmpname = fvp->v_name;
fvp->v_name = svp->v_name;
svp->v_name = tmpname;
if (fvp->v_parent != svp->v_parent) {
vnode_t tmp;
tmp = fvp->v_parent;
fvp->v_parent = svp->v_parent;
svp->v_parent = tmp;
}
name_cache_unlock();
#if CONFIG_FSE
if (fpath != NULL && spath != NULL) {
add_fsevent(FSE_EXCHANGE, ctx,
FSE_ARG_STRING, flen, fpath,
FSE_ARG_FINFO, &f_finfo,
FSE_ARG_STRING, slen, spath,
FSE_ARG_FINFO, &s_finfo,
FSE_ARG_DONE);
}
#endif
}
out:
if (fpath != NULL) {
RELEASE_PATH(fpath);
}
if (spath != NULL) {
RELEASE_PATH(spath);
}
vnode_put(svp);
vnode_put(fvp);
out2:
return error;
}
/*
* Return (in MB) the amount of freespace on the given vnode's volume.
*/
uint32_t freespace_mb(vnode_t vp);
uint32_t
freespace_mb(vnode_t vp)
{
vfs_update_vfsstat(vp->v_mount, vfs_context_current(), VFS_USER_EVENT);
return (uint32_t)(((uint64_t)vp->v_mount->mnt_vfsstat.f_bavail *
vp->v_mount->mnt_vfsstat.f_bsize) >> 20);
}
#if CONFIG_SEARCHFS
/* ARGSUSED */
int
searchfs(proc_t p, struct searchfs_args *uap, __unused int32_t *retval)
{
vnode_t vp, tvp;
int i, error = 0;
int fserror = 0;
struct nameidata nd;
struct user64_fssearchblock searchblock;
struct searchstate *state;
struct attrlist *returnattrs;
struct timeval timelimit;
void *searchparams1, *searchparams2;
uio_t auio = NULL;
int spacetype = proc_is64bit(p) ? UIO_USERSPACE64 : UIO_USERSPACE32;
uint32_t nummatches;
size_t mallocsize;
uint32_t nameiflags;
vfs_context_t ctx = vfs_context_current();
UIO_STACKBUF(uio_buf, 1);
/* Start by copying in fsearchblock parameter list */
if (IS_64BIT_PROCESS(p)) {
error = copyin(uap->searchblock, (caddr_t) &searchblock, sizeof(searchblock));
timelimit.tv_sec = searchblock.timelimit.tv_sec;
timelimit.tv_usec = searchblock.timelimit.tv_usec;
} else {
struct user32_fssearchblock tmp_searchblock;
error = copyin(uap->searchblock, (caddr_t) &tmp_searchblock, sizeof(tmp_searchblock));
// munge into 64-bit version
searchblock.returnattrs = CAST_USER_ADDR_T(tmp_searchblock.returnattrs);
searchblock.returnbuffer = CAST_USER_ADDR_T(tmp_searchblock.returnbuffer);
searchblock.returnbuffersize = tmp_searchblock.returnbuffersize;
searchblock.maxmatches = tmp_searchblock.maxmatches;
/*
* These casts are safe. We will promote the tv_sec into a 64 bit long if necessary
* from a 32 bit long, and tv_usec is already a signed 32 bit int.
*/
timelimit.tv_sec = (__darwin_time_t) tmp_searchblock.timelimit.tv_sec;
timelimit.tv_usec = (__darwin_useconds_t) tmp_searchblock.timelimit.tv_usec;
searchblock.searchparams1 = CAST_USER_ADDR_T(tmp_searchblock.searchparams1);
searchblock.sizeofsearchparams1 = tmp_searchblock.sizeofsearchparams1;
searchblock.searchparams2 = CAST_USER_ADDR_T(tmp_searchblock.searchparams2);
searchblock.sizeofsearchparams2 = tmp_searchblock.sizeofsearchparams2;
searchblock.searchattrs = tmp_searchblock.searchattrs;
}
if (error) {
return error;
}
/* Do a sanity check on sizeofsearchparams1 and sizeofsearchparams2.
*/
if (searchblock.sizeofsearchparams1 > SEARCHFS_MAX_SEARCHPARMS ||
searchblock.sizeofsearchparams2 > SEARCHFS_MAX_SEARCHPARMS) {
return EINVAL;
}
/* Now malloc a big bunch of space to hold the search parameters, the attrlists and the search state. */
/* It all has to do into local memory and it's not that big so we might as well put it all together. */
/* Searchparams1 shall be first so we might as well use that to hold the base address of the allocated*/
/* block. */
/* */
/* NOTE: we allocate an extra 8 bytes to account for the difference in size of the searchstate */
/* due to the changes in rdar://problem/12438273. That way if a 3rd party file system */
/* assumes the size is still 556 bytes it will continue to work */
mallocsize = searchblock.sizeofsearchparams1 + searchblock.sizeofsearchparams2 +
sizeof(struct attrlist) + sizeof(struct searchstate) + (2 * sizeof(uint32_t));
searchparams1 = kalloc_data(mallocsize, Z_WAITOK);
/* Now set up the various pointers to the correct place in our newly allocated memory */
searchparams2 = (void *) (((caddr_t) searchparams1) + searchblock.sizeofsearchparams1);
returnattrs = (struct attrlist *) (((caddr_t) searchparams2) + searchblock.sizeofsearchparams2);
state = (struct searchstate *) (((caddr_t) returnattrs) + sizeof(struct attrlist));
/* Now copy in the stuff given our local variables. */
if ((error = copyin(searchblock.searchparams1, searchparams1, searchblock.sizeofsearchparams1))) {
goto freeandexit;
}
if ((error = copyin(searchblock.searchparams2, searchparams2, searchblock.sizeofsearchparams2))) {
goto freeandexit;
}
if ((error = copyin(searchblock.returnattrs, (caddr_t) returnattrs, sizeof(struct attrlist)))) {
goto freeandexit;
}
if ((error = copyin(uap->state, (caddr_t) state, sizeof(struct searchstate)))) {
goto freeandexit;
}
/*
* When searching a union mount, need to set the
* start flag at the first call on each layer to
* reset state for the new volume.
*/
if (uap->options & SRCHFS_START) {
state->ss_union_layer = 0;
} else {
uap->options |= state->ss_union_flags;
}
state->ss_union_flags = 0;
/*
* Because searchparams1 and searchparams2 may contain an ATTR_CMN_NAME search parameter,
* which is passed in with an attrreference_t, we need to inspect the buffer manually here.
* The KPI does not provide us the ability to pass in the length of the buffers searchparams1
* and searchparams2. To obviate the need for all searchfs-supporting filesystems to
* validate the user-supplied data offset of the attrreference_t, we'll do it here.
*/
if (searchblock.searchattrs.commonattr & ATTR_CMN_NAME) {
attrreference_t* string_ref;
u_int32_t* start_length;
user64_size_t param_length;
/* validate searchparams1 */
param_length = searchblock.sizeofsearchparams1;
/* skip the word that specifies length of the buffer */
start_length = (u_int32_t*) searchparams1;
start_length = start_length + 1;
string_ref = (attrreference_t*) start_length;
/* ensure no negative offsets or too big offsets */
if (string_ref->attr_dataoffset < 0) {
error = EINVAL;
goto freeandexit;
}
if (string_ref->attr_length > MAXPATHLEN) {
error = EINVAL;
goto freeandexit;
}
/* Check for pointer overflow in the string ref */
if (((char*) string_ref + string_ref->attr_dataoffset) < (char*) string_ref) {
error = EINVAL;
goto freeandexit;
}
if (((char*) string_ref + string_ref->attr_dataoffset) > ((char*)searchparams1 + param_length)) {
error = EINVAL;
goto freeandexit;
}
if (((char*)string_ref + string_ref->attr_dataoffset + string_ref->attr_length) > ((char*)searchparams1 + param_length)) {
error = EINVAL;
goto freeandexit;
}
}
/* set up the uio structure which will contain the users return buffer */
auio = uio_createwithbuffer(1, 0, spacetype, UIO_READ, &uio_buf[0], sizeof(uio_buf));
uio_addiov(auio, searchblock.returnbuffer, searchblock.returnbuffersize);
nameiflags = 0;
if ((uap->options & FSOPT_NOFOLLOW) == 0) {
nameiflags |= FOLLOW;
}
NDINIT(&nd, LOOKUP, OP_SEARCHFS, nameiflags | AUDITVNPATH1,
UIO_USERSPACE, uap->path, ctx);
error = namei(&nd);
if (error) {
goto freeandexit;
}
vp = nd.ni_vp;
nameidone(&nd);
/*
* Switch to the root vnode for the volume
*/
error = VFS_ROOT(vnode_mount(vp), &tvp, ctx);
vnode_put(vp);
if (error) {
goto freeandexit;
}
vp = tvp;
#if CONFIG_UNION_MOUNTS
/*
* If it's a union mount, the path lookup takes
* us to the top layer. But we may need to descend
* to a lower layer. For non-union mounts the layer
* is always zero.
*/
for (i = 0; i < (int) state->ss_union_layer; i++) {
if ((vp->v_mount->mnt_flag & MNT_UNION) == 0) {
break;
}
tvp = vp;
vp = vp->v_mount->mnt_vnodecovered;
if (vp == NULL) {
vnode_put(tvp);
error = ENOENT;
goto freeandexit;
}
error = vnode_getwithref(vp);
vnode_put(tvp);
if (error) {
goto freeandexit;
}
}
#endif /* CONFIG_UNION_MOUNTS */
#if CONFIG_MACF
error = mac_vnode_check_searchfs(ctx, vp, returnattrs, &searchblock.searchattrs);
if (error) {
vnode_put(vp);
goto freeandexit;
}
#endif
/*
* If searchblock.maxmatches == 0, then skip the search. This has happened
* before and sometimes the underlying code doesnt deal with it well.
*/
if (searchblock.maxmatches == 0) {
nummatches = 0;
goto saveandexit;
}
/*
* Allright, we have everything we need, so lets make that call.
*
* We keep special track of the return value from the file system:
* EAGAIN is an acceptable error condition that shouldn't keep us
* from copying out any results...
*/
fserror = VNOP_SEARCHFS(vp,
searchparams1,
searchparams2,
&searchblock.searchattrs,
(uint32_t)searchblock.maxmatches,
&timelimit,
returnattrs,
&nummatches,
(uint32_t)uap->scriptcode,
(uint32_t)uap->options,
auio,
(struct searchstate *) &state->ss_fsstate,
ctx);
#if CONFIG_UNION_MOUNTS
/*
* If it's a union mount we need to be called again
* to search the mounted-on filesystem.
*/
if ((vp->v_mount->mnt_flag & MNT_UNION) && fserror == 0) {
state->ss_union_flags = SRCHFS_START;
state->ss_union_layer++; // search next layer down
fserror = EAGAIN;
}
#endif /* CONFIG_UNION_MOUNTS */
saveandexit:
vnode_put(vp);
/* Now copy out the stuff that needs copying out. That means the number of matches, the
* search state. Everything was already put into he return buffer by the vop call. */
if ((error = copyout((caddr_t) state, uap->state, sizeof(struct searchstate))) != 0) {
goto freeandexit;
}
if ((error = suulong(uap->nummatches, (uint64_t)nummatches)) != 0) {
goto freeandexit;
}
error = fserror;
freeandexit:
kfree_data(searchparams1, mallocsize);
return error;
} /* end of searchfs system call */
#else /* CONFIG_SEARCHFS */
int
searchfs(__unused proc_t p, __unused struct searchfs_args *uap, __unused int32_t *retval)
{
return ENOTSUP;
}
#endif /* CONFIG_SEARCHFS */
#if CONFIG_DATALESS_FILES
/*
* === Namespace Resolver Up-call Mechanism ===
*
* When I/O is performed to a dataless file or directory (read, write,
* lookup-in, etc.), the file system performs an upcall to the namespace
* resolver (filecoordinationd) to materialize the object.
*
* We need multiple up-calls to be in flight at once, and we need these
* up-calls to be interruptible, thus the following implementation:
*
* => The nspace_resolver_request represents the in-kernel request state.
* It contains a request ID, storage space for the errno code returned
* by filecoordinationd, and flags.
*
* => The request ID is simply a global monotonically incrementing 32-bit
* number. Outstanding requests are stored in a hash table, and the
* hash function is extremely simple.
*
* => When an upcall is to be made to filecoordinationd, a request structure
* is allocated on the stack (it is small, and needs to live only during
* the duration of the call to resolve_nspace_item_ext()). It is
* initialized and inserted into the table. Some backpressure from
* filecoordinationd is applied by limiting the numnber of entries that
* can be inserted into the table (and thus limiting the number of
* outstanding requests issued to filecoordinationd); waiting for an
* available slot is interruptible.
*
* => Once the request has been inserted into the table, the up-call is made
* to filecoordinationd via a MiG-generated stub. The up-call returns
* immediately and filecoordinationd processes the request asynchronously.
*
* => The caller now waits for the request to complete. Tnis is achieved by
* sleeping on the address of the request structure and waiting for
* filecoordinationd to mark the request structure as complete. This
* is an interruptible sleep call; if interrupted, the request structure
* is removed from the table and EINTR is returned to the caller. If
* this occurs, an advisory up-call is made to filecoordinationd with
* the request ID to indicate that the request can be aborted or
* de-prioritized at the discretion of filecoordinationd.
*
* => When filecoordinationd has completed the request, it signals completion
* by writing to the vfs.nspace.complete sysctl node. Only a process
* decorated as a namespace resolver can write to this sysctl node. The
* value is a request ID / errno tuple passed as an array of 2 uint32_t's.
* The request ID is looked up in the table, and if the request is found,
* the error code is stored in the request structure and a wakeup()
* issued on the address of the request structure. If the request is not
* found, we simply drop the completion notification, assuming that the
* caller was interrupted.
*
* => When the waiting thread wakes up, it extracts the error code from the
* request structure, removes the request from the table, and returns the
* error code to the calling function. Fini!
*/
struct nspace_resolver_request {
LIST_ENTRY(nspace_resolver_request) r_hashlink;
vnode_t r_vp;
vnode_t r_tdvp;
uint32_t r_req_id;
int r_resolver_error;
int r_flags;
};
#define RRF_COMPLETE 0x0001
#define RRF_COMPLETING 0x0002
struct nspace_resolver_completion_data {
uint32_t req_id;
int32_t resolver_error;
uint64_t orig_gencount;
uint64_t orig_syncroot;
};
static uint32_t
next_nspace_req_id(void)
{
static uint32_t next_req_id;
return OSAddAtomic(1, &next_req_id);
}
#define NSPACE_RESOLVER_REQ_HASHSIZE 32 /* XXX tune */
#define NSPACE_RESOLVER_MAX_OUTSTANDING 256 /* XXX tune */
static LIST_HEAD(nspace_resolver_requesthead,
nspace_resolver_request) * nspace_resolver_request_hashtbl;
static u_long nspace_resolver_request_hashmask;
static u_int nspace_resolver_request_count;
static bool nspace_resolver_request_wait_slot;
static LCK_GRP_DECLARE(nspace_resolver_request_lck_grp, "file namespace resolver");
static LCK_MTX_DECLARE(nspace_resolver_request_hash_mutex,
&nspace_resolver_request_lck_grp);
#define NSPACE_REQ_LOCK() \
lck_mtx_lock(&nspace_resolver_request_hash_mutex)
#define NSPACE_REQ_UNLOCK() \
lck_mtx_unlock(&nspace_resolver_request_hash_mutex)
#define NSPACE_RESOLVER_HASH(req_id) \
(&nspace_resolver_request_hashtbl[(req_id) & \
nspace_resolver_request_hashmask])
static struct nspace_resolver_request *
nspace_resolver_req_lookup(uint32_t req_id, bool skip_completing)
{
struct nspace_resolver_requesthead *bucket;
struct nspace_resolver_request *req;
bucket = NSPACE_RESOLVER_HASH(req_id);
LIST_FOREACH(req, bucket, r_hashlink) {
if (req->r_req_id == req_id) {
/*
* If this request already has a completion
* pending, don't return it again.
*/
if ((req->r_flags & RRF_COMPLETING) != 0 &&
skip_completing) {
req = NULL;
}
return req;
}
}
return NULL;
}
static int
nspace_resolver_req_add(struct nspace_resolver_request *req)
{
struct nspace_resolver_requesthead *bucket;
int error;
NSPACE_REQ_LOCK();
while (nspace_resolver_request_count >=
NSPACE_RESOLVER_MAX_OUTSTANDING) {
nspace_resolver_request_wait_slot = true;
error = msleep(&nspace_resolver_request_count,
&nspace_resolver_request_hash_mutex,
PVFS | PCATCH, "nspacerq", NULL);
if (error) {
NSPACE_REQ_UNLOCK();
return error;
}
}
bucket = NSPACE_RESOLVER_HASH(req->r_req_id);
#if DIAGNOSTIC
assert(nspace_resolver_req_lookup(req->r_req_id, false) == NULL);
#endif /* DIAGNOSTIC */
LIST_INSERT_HEAD(bucket, req, r_hashlink);
nspace_resolver_request_count++;
NSPACE_REQ_UNLOCK();
return 0;
}
static void
nspace_resolver_req_wait_pending_completion(struct nspace_resolver_request *req)
{
/*
* If a completion is in-progress, we have to wait for the
* completion handler to finish because it's still using 'req',
* which is allocated on our stack a couple of frames up.
*/
while ((req->r_flags & RRF_COMPLETING) != 0) {
(void) msleep(req, &nspace_resolver_request_hash_mutex,
PVFS, "nspacecmplt", NULL);
}
}
static void
nspace_resolver_req_remove_and_unlock(struct nspace_resolver_request *req)
{
struct nspace_resolver_requesthead *bucket;
/* We're called with NSPACE_REQ_LOCK held. */
bucket = NSPACE_RESOLVER_HASH(req->r_req_id);
#if DIAGNOSTIC
assert((req->r_flags & RRF_COMPLETING) == 0);
assert(nspace_resolver_req_lookup(req->r_req_id, false) != NULL);
#endif /* DIAGNOSTIC */
LIST_REMOVE(req, r_hashlink);
nspace_resolver_request_count--;
if (nspace_resolver_request_wait_slot) {
nspace_resolver_request_wait_slot = false;
wakeup(&nspace_resolver_request_count);
}
nspace_resolver_req_wait_pending_completion(req);
NSPACE_REQ_UNLOCK();
}
static void
nspace_resolver_req_remove(struct nspace_resolver_request *req)
{
NSPACE_REQ_LOCK();
nspace_resolver_req_remove_and_unlock(req);
}
static void
nspace_resolver_req_cancel(uint32_t req_id)
{
kern_return_t kr;
mach_port_t mp;
// Failures here aren't fatal -- the cancellation message
// sent to the resolver is merely advisory.
kr = host_get_filecoordinationd_port(host_priv_self(), &mp);
if (kr != KERN_SUCCESS || !IPC_PORT_VALID(mp)) {
return;
}
kr = send_nspace_resolve_cancel(mp, req_id);
if (kr != KERN_SUCCESS) {
os_log_error(OS_LOG_DEFAULT,
"NSPACE send_nspace_resolve_cancel failure: %d", kr);
}
ipc_port_release_send(mp);
}
static int
nspace_resolver_req_wait(struct nspace_resolver_request *req)
{
bool send_cancel_message = false;
int error;
NSPACE_REQ_LOCK();
while ((req->r_flags & RRF_COMPLETE) == 0) {
error = msleep(req, &nspace_resolver_request_hash_mutex,
PVFS | PCATCH, "nspace", NULL);
if (error && error != ERESTART) {
req->r_resolver_error = (error == EINTR) ? EINTR :
ETIMEDOUT;
send_cancel_message = true;
break;
}
}
nspace_resolver_req_remove_and_unlock(req);
/*
* It's safe to continue referencing 'req' here because it's
* allocated on our caller's stack.
*/
if (send_cancel_message) {
nspace_resolver_req_cancel(req->r_req_id);
}
return req->r_resolver_error;
}
static void
nspace_resolver_req_mark_complete(
struct nspace_resolver_request *req,
int resolver_error)
{
req->r_resolver_error = resolver_error;
req->r_flags = (req->r_flags & ~RRF_COMPLETING) | RRF_COMPLETE;
wakeup(req);
}
static void
nspace_resolver_req_mark_completion_pending(struct nspace_resolver_request *req)
{
req->r_flags |= RRF_COMPLETING;
}
static void
nspace_resolver_req_completed(const struct nspace_resolver_completion_data *c)
{
struct nspace_resolver_request *req;
int error;
struct vnode_attr va;
vnode_t vp;
NSPACE_REQ_LOCK();
req = nspace_resolver_req_lookup(c->req_id, true);
if (req == NULL) {
/*
* If we don't find the request corresponding to our req_id,
* just drop the completion on the floor; it's likely that
* the requester interrupted with a signal, or it may already
* be completing.
*/
NSPACE_REQ_UNLOCK();
return;
}
/*
* Get out now if the resolver reported an error.
*/
if ((error = c->resolver_error) != 0) {
goto out;
}
/*
* If the resolver did not specify any namespace shape criteria
* for letting the operation proceed, then get out now.
*/
if (c->orig_gencount == 0 && c->orig_syncroot == 0) {
goto out;
}
/*
* We're going to have to acquire the mount rename lock and do
* some I/O in order to verify the criteria. Mark the request
* as pending so no one else messes with it after we drop the
* NSPACE_REQ_LOCK.
*/
nspace_resolver_req_mark_completion_pending(req);
NSPACE_REQ_UNLOCK();
/*
* Lock out renames from changing the shape of the tree while
* validate the criteria.
*/
mount_t locked_mp = req->r_vp->v_mount;
mount_ref(locked_mp, 0);
mount_lock_renames(locked_mp);
if (c->orig_gencount != 0) {
vp = req->r_vp;
if (error) {
goto out_dropmount;
}
VATTR_INIT(&va);
VATTR_WANTED(&va, va_recursive_gencount);
error = vnode_getattr(vp, &va, vfs_context_kernel());
if (error) {
goto out_dropmount;
}
if (VATTR_NOT_RETURNED(&va, va_recursive_gencount) ||
va.va_recursive_gencount != c->orig_gencount) {
printf("nspace.complete: gencount changed! (orig %llu cur %llu)\n",
c->orig_gencount, va.va_recursive_gencount);
error = EBUSY;
goto out_dropmount;
}
}
/*
* Ignore orig_syncroot if a destination directory wasn't specified
* in the request.
*/
if (c->orig_syncroot != 0 && (vp = req->r_tdvp) != NULL) {
uint64_t syncroot_id;
if (error) {
goto out_dropmount;
}
#ifndef APFSIOC_GET_SYNC_ROOT
#define APFSIOC_GET_SYNC_ROOT _IOR('J', 115, uint64_t)
#endif
error = VNOP_IOCTL(vp, APFSIOC_GET_SYNC_ROOT,
(caddr_t)&syncroot_id, 0, vfs_context_kernel());
if (error) {
goto out_dropmount;
}
if (syncroot_id != c->orig_syncroot) {
printf("nspace.complete: syncroot changed! (orig %llu cur %llu)\n",
c->orig_syncroot, syncroot_id);
error = EBUSY;
goto out_dropmount;
}
}
out_dropmount:
mount_unlock_renames(locked_mp);
mount_drop(locked_mp, 0);
NSPACE_REQ_LOCK();
out:
nspace_resolver_req_mark_complete(req, error);
NSPACE_REQ_UNLOCK();
}
static struct proc *nspace_resolver_proc;
static int
nspace_resolver_get_proc_state(struct proc *p, int *is_resolver)
{
*is_resolver = ((p->p_lflag & P_LNSPACE_RESOLVER) &&
p == nspace_resolver_proc) ? 1 : 0;
return 0;
}
static boolean_t vfs_context_is_dataless_resolver(vfs_context_t);
static int
nspace_resolver_set_proc_state(struct proc *p, int is_resolver)
{
vfs_context_t ctx = vfs_context_current();
int error = 0;
//
// The system filecoordinationd runs as uid == 0. This also
// has the nice side-effect of filtering out filecoordinationd
// running in the simulator.
//
if (!vfs_context_issuser(ctx) ||
!vfs_context_is_dataless_resolver(ctx)) {
return EPERM;
}
if (is_resolver) {
NSPACE_REQ_LOCK();
if (nspace_resolver_proc == NULL) {
proc_lock(p);
p->p_lflag |= P_LNSPACE_RESOLVER;
proc_unlock(p);
nspace_resolver_proc = p;
} else {
error = EBUSY;
}
NSPACE_REQ_UNLOCK();
} else {
// This is basically just like the exit case.
// nspace_resolver_exited() will verify that the
// process is the resolver, and will clear the
// global.
nspace_resolver_exited(p);
}
return error;
}
static int
nspace_materialization_get_proc_state(struct proc *p, int *is_prevented)
{
if ((p->p_lflag & P_LNSPACE_RESOLVER) != 0 ||
(p->p_vfs_iopolicy &
P_VFS_IOPOLICY_MATERIALIZE_DATALESS_FILES) == 0) {
*is_prevented = 1;
} else {
*is_prevented = 0;
}
return 0;
}
static int
nspace_materialization_set_proc_state(struct proc *p, int is_prevented)
{
if (p->p_lflag & P_LNSPACE_RESOLVER) {
return is_prevented ? 0 : EBUSY;
}
if (is_prevented) {
OSBitAndAtomic16(~((uint16_t)P_VFS_IOPOLICY_MATERIALIZE_DATALESS_FILES), &p->p_vfs_iopolicy);
} else {
OSBitOrAtomic16((uint16_t)P_VFS_IOPOLICY_MATERIALIZE_DATALESS_FILES, &p->p_vfs_iopolicy);
}
return 0;
}
static int
nspace_materialization_get_thread_state(int *is_prevented)
{
uthread_t ut = current_uthread();
*is_prevented = (ut->uu_flag & UT_NSPACE_NODATALESSFAULTS) ? 1 : 0;
return 0;
}
static int
nspace_materialization_set_thread_state(int is_prevented)
{
uthread_t ut = current_uthread();
if (is_prevented) {
ut->uu_flag |= UT_NSPACE_NODATALESSFAULTS;
} else {
ut->uu_flag &= ~UT_NSPACE_NODATALESSFAULTS;
}
return 0;
}
/* the vfs.nspace branch */
SYSCTL_NODE(_vfs, OID_AUTO, nspace, CTLFLAG_RW | CTLFLAG_LOCKED, NULL, "vfs nspace hinge");
static int
sysctl_nspace_resolver(__unused struct sysctl_oid *oidp,
__unused void *arg1, __unused int arg2, struct sysctl_req *req)
{
struct proc *p = req->p;
int new_value, old_value, changed = 0;
int error;
error = nspace_resolver_get_proc_state(p, &old_value);
if (error) {
return error;
}
error = sysctl_io_number(req, old_value, sizeof(int), &new_value,
&changed);
if (error == 0 && changed) {
error = nspace_resolver_set_proc_state(p, new_value);
}
return error;
}
/* decorate this process as the dataless file resolver */
SYSCTL_PROC(_vfs_nspace, OID_AUTO, resolver,
CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_ANYBODY | CTLFLAG_LOCKED,
0, 0, sysctl_nspace_resolver, "I", "");
static int
sysctl_nspace_prevent_materialization(__unused struct sysctl_oid *oidp,
__unused void *arg1, __unused int arg2, struct sysctl_req *req)
{
struct proc *p = req->p;
int new_value, old_value, changed = 0;
int error;
error = nspace_materialization_get_proc_state(p, &old_value);
if (error) {
return error;
}
error = sysctl_io_number(req, old_value, sizeof(int), &new_value,
&changed);
if (error == 0 && changed) {
error = nspace_materialization_set_proc_state(p, new_value);
}
return error;
}
/* decorate this process as not wanting to materialize dataless files */
SYSCTL_PROC(_vfs_nspace, OID_AUTO, prevent_materialization,
CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_ANYBODY | CTLFLAG_LOCKED,
0, 0, sysctl_nspace_prevent_materialization, "I", "");
static int
sysctl_nspace_thread_prevent_materialization(__unused struct sysctl_oid *oidp,
__unused void *arg1, __unused int arg2, struct sysctl_req *req)
{
int new_value, old_value, changed = 0;
int error;
error = nspace_materialization_get_thread_state(&old_value);
if (error) {
return error;
}
error = sysctl_io_number(req, old_value, sizeof(int), &new_value,
&changed);
if (error == 0 && changed) {
error = nspace_materialization_set_thread_state(new_value);
}
return error;
}
/* decorate this thread as not wanting to materialize dataless files */
SYSCTL_PROC(_vfs_nspace, OID_AUTO, thread_prevent_materialization,
CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_ANYBODY | CTLFLAG_LOCKED,
0, 0, sysctl_nspace_thread_prevent_materialization, "I", "");
static int
sysctl_nspace_complete(__unused struct sysctl_oid *oidp, __unused void *arg1,
__unused int arg2, struct sysctl_req *req)
{
struct proc *p = req->p;
uint32_t req_status[2] = { 0, 0 };
uint64_t gencount = 0;
uint64_t syncroot = 0;
int error, is_resolver, changed = 0, other_changed;
error = nspace_resolver_get_proc_state(p, &is_resolver);
if (error) {
return error;
}
if (!is_resolver) {
return EPERM;
}
error = sysctl_io_opaque(req, req_status, sizeof(req_status),
&changed);
if (error) {
return error;
}
/*
* Get the gencount if it was passed. Ignore errors, because
* it's optional.
*/
error = sysctl_io_opaque(req, &gencount, sizeof(gencount),
&other_changed);
if (error) {
gencount = 0;
error = 0;
}
/*
* ...and now the syncroot ID.
*/
error = sysctl_io_opaque(req, &syncroot, sizeof(syncroot),
&other_changed);
if (error) {
syncroot = 0;
error = 0;
}
/*
* req_status[0] is the req_id
*
* req_status[1] is the errno
*/
if (error == 0 && changed) {
const struct nspace_resolver_completion_data cd = {
.req_id = req_status[0],
.resolver_error = req_status[1],
.orig_gencount = gencount,
.orig_syncroot = syncroot,
};
nspace_resolver_req_completed(&cd);
}
return error;
}
/* Resolver reports completed reqs here. */
SYSCTL_PROC(_vfs_nspace, OID_AUTO, complete,
CTLTYPE_OPAQUE | CTLFLAG_RW | CTLFLAG_ANYBODY | CTLFLAG_LOCKED,
0, 0, sysctl_nspace_complete, "-", "");
#endif /* CONFIG_DATALESS_FILES */
#if CONFIG_DATALESS_FILES
#define __no_dataless_unused /* nothing */
#else
#define __no_dataless_unused __unused
#endif
int
vfs_context_dataless_materialization_is_prevented(
vfs_context_t const ctx __no_dataless_unused)
{
#if CONFIG_DATALESS_FILES
proc_t const p = vfs_context_proc(ctx);
thread_t const t = vfs_context_thread(ctx);
uthread_t const ut = t ? get_bsdthread_info(t) : NULL;
/*
* Kernel context ==> return EDEADLK, as we would with any random
* process decorated as no-materialize.
*/
if (ctx == vfs_context_kernel()) {
return EDEADLK;
}
/*
* If the process has the dataless-manipulation entitlement,
* materialization is prevented, and depending on the kind
* of file system operation, things get to proceed as if the
* object is not dataless.
*/
if (vfs_context_is_dataless_manipulator(ctx)) {
return EJUSTRETURN;
}
/*
* Per-thread decorations override any process-wide decorations.
* (Foundation uses this, and this overrides even the dataless-
* manipulation entitlement so as to make API contracts consistent.)
*/
if (ut != NULL) {
if (ut->uu_flag & UT_NSPACE_NODATALESSFAULTS) {
return EDEADLK;
}
if (ut->uu_flag & UT_NSPACE_FORCEDATALESSFAULTS) {
return 0;
}
}
/*
* If the process's iopolicy specifies that dataless files
* can be materialized, then we let it go ahead.
*/
if (p->p_vfs_iopolicy & P_VFS_IOPOLICY_MATERIALIZE_DATALESS_FILES) {
return 0;
}
#endif /* CONFIG_DATALESS_FILES */
/*
* The default behavior is to not materialize dataless files;
* return to the caller that deadlock was detected.
*/
return EDEADLK;
}
void
nspace_resolver_init(void)
{
#if CONFIG_DATALESS_FILES
nspace_resolver_request_hashtbl =
hashinit(NSPACE_RESOLVER_REQ_HASHSIZE,
M_VNODE /* XXX */, &nspace_resolver_request_hashmask);
#endif /* CONFIG_DATALESS_FILES */
}
void
nspace_resolver_exited(struct proc *p __no_dataless_unused)
{
#if CONFIG_DATALESS_FILES
struct nspace_resolver_requesthead *bucket;
struct nspace_resolver_request *req;
u_long idx;
NSPACE_REQ_LOCK();
if ((p->p_lflag & P_LNSPACE_RESOLVER) &&
p == nspace_resolver_proc) {
for (idx = 0; idx <= nspace_resolver_request_hashmask; idx++) {
bucket = &nspace_resolver_request_hashtbl[idx];
LIST_FOREACH(req, bucket, r_hashlink) {
nspace_resolver_req_wait_pending_completion(req);
nspace_resolver_req_mark_complete(req,
ETIMEDOUT);
}
}
nspace_resolver_proc = NULL;
}
NSPACE_REQ_UNLOCK();
#endif /* CONFIG_DATALESS_FILES */
}
#define DATALESS_RESOLVER_ENTITLEMENT \
"com.apple.private.vfs.dataless-resolver"
#define DATALESS_MANIPULATION_ENTITLEMENT \
"com.apple.private.vfs.dataless-manipulation"
#if CONFIG_DATALESS_FILES
/*
* Return TRUE if the vfs context is associated with the dataless
* resolver.
*/
static boolean_t
vfs_context_is_dataless_resolver(vfs_context_t ctx __no_dataless_unused)
{
return IOTaskHasEntitlement(vfs_context_task(ctx),
DATALESS_RESOLVER_ENTITLEMENT);
}
#endif /* CONFIG_DATALESS_FILES */
/*
* Return TRUE if the vfs context is associated with a process entitled
* for dataless manipulation.
*
* XXX Arguably belongs in vfs_subr.c, but is here because of the
* complication around CONFIG_DATALESS_FILES.
*/
boolean_t
vfs_context_is_dataless_manipulator(vfs_context_t ctx __no_dataless_unused)
{
#if CONFIG_DATALESS_FILES
task_t task = vfs_context_task(ctx);
return IOTaskHasEntitlement(task, DATALESS_MANIPULATION_ENTITLEMENT) ||
IOTaskHasEntitlement(task, DATALESS_RESOLVER_ENTITLEMENT);
#else
return false;
#endif /* CONFIG_DATALESS_FILES */
}
#if CONFIG_DATALESS_FILES
static void
log_materialization_prevented(vnode_t vp, uint64_t op)
{
char p_name[MAXCOMLEN + 1];
char *vntype;
proc_selfname(&p_name[0], sizeof(p_name));
if (vp->v_type == VREG) {
vntype = "File";
} else if (vp->v_type == VDIR) {
vntype = "Dir";
} else if (vp->v_type == VLNK) {
vntype = "SymLink";
} else {
vntype = "Other";
}
#if DEVELOPMENT
char *path = NULL;
int len;
path = get_pathbuff();
len = MAXPATHLEN;
if (path) {
vn_getpath(vp, path, &len);
}
os_log_debug(OS_LOG_DEFAULT,
"NSPACE process %s (pid %d) is decorated as no-materialization (op %lld; %s) path: %s",
p_name, proc_selfpid(),
op, vntype, path ? path : "<unknown-path>");
if (path) {
release_pathbuff(path);
}
#else
os_log_debug(OS_LOG_DEFAULT,
"NSPACE process %s (pid %d) is decorated as no-materialization (op %lld; %s)",
p_name, proc_selfpid(),
op, vntype);
#endif
}
#endif /* CONFIG_DATALESS_FILES */
static int
vfs_materialize_item(
vnode_t vp __no_dataless_unused,
uint32_t op __no_dataless_unused,
int64_t offset __no_dataless_unused,
int64_t size __no_dataless_unused,
char *lookup_name __no_dataless_unused,
size_t const namelen __no_dataless_unused,
vnode_t tdvp __no_dataless_unused)
{
#if CONFIG_DATALESS_FILES
kern_return_t kern_ret;
mach_port_t mach_port;
char *path = NULL;
vfs_context_t context;
int path_len;
int error;
audit_token_t atoken;
enum vtype vp_vtype;
/* Swap files are special; ignore them */
if (vnode_isswap(vp)) {
return 0;
}
/*
* NAMESPACE_HANDLER_SNAPSHOT_EVENT and NAMESPACE_HANDLER_TRACK_EVENT
* are no longer used nor supported.
*/
if (op & NAMESPACE_HANDLER_SNAPSHOT_EVENT) {
os_log_debug(OS_LOG_DEFAULT, "NSPACE SNAPSHOT not handled");
return ENOTSUP;
}
if (op & NAMESPACE_HANDLER_TRACK_EVENT) {
os_log_debug(OS_LOG_DEFAULT, "NSPACE TRACK not handled");
return ENOTSUP;
}
/* Normalize 'op'. */
op &= ~NAMESPACE_HANDLER_EVENT_TYPE_MASK;
/*
* To-directory is only meaningful for rename operations;
* ignore it if someone handed one to us unexpectedly.
*/
if (op != NAMESPACE_HANDLER_RENAME_OP) {
tdvp = NULL;
}
context = vfs_context_current();
/* Remember this for later. */
vp_vtype = vnode_vtype(vp);
error = vfs_context_dataless_materialization_is_prevented(context);
if (error) {
log_materialization_prevented(vp, op);
goto out_check_errors;
}
kern_ret = host_get_filecoordinationd_port(host_priv_self(),
&mach_port);
if (kern_ret != KERN_SUCCESS || !IPC_PORT_VALID(mach_port)) {
os_log_error(OS_LOG_DEFAULT, "NSPACE no port");
/*
* Treat this like being unable to access the backing store
* server.
*/
return ETIMEDOUT;
}
int path_alloc_len = MAXPATHLEN;
do {
path = kalloc_data(path_alloc_len, Z_WAITOK | Z_ZERO);
if (path == NULL) {
return ENOMEM;
}
path_len = path_alloc_len;
error = vn_getpath(vp, path, &path_len);
if (error == 0) {
break;
} else if (error == ENOSPC) {
kfree_data(path, path_alloc_len);
path = NULL;
} else {
goto out_release_port;
}
} while (error == ENOSPC && (path_alloc_len += MAXPATHLEN) && path_alloc_len <= FSGETPATH_MAXBUFLEN);
error = vfs_context_copy_audit_token(context, &atoken);
if (error) {
goto out_release_port;
}
struct nspace_resolver_request req = {
.r_req_id = next_nspace_req_id(),
.r_vp = vp,
.r_tdvp = tdvp,
};
error = nspace_resolver_req_add(&req);
if (error) {
goto out_release_port;
}
os_log_debug(OS_LOG_DEFAULT, "NSPACE resolve_path call");
if (op == NAMESPACE_HANDLER_RENAME_OP && tdvp != NULL) {
char *dest_path = NULL;
int dest_path_len;
dest_path = zalloc(ZV_NAMEI);
dest_path_len = MAXPATHLEN;
error = vn_getpath(tdvp, dest_path, &dest_path_len);
if (error) {
zfree(ZV_NAMEI, dest_path);
goto out_release_port;
}
/*
* Force setting NAMESPACE_HANDLER_NSPACE_EVENT for
* compatibility with existing agents in user-space
* who get passed this value.
*/
kern_ret = send_vfs_resolve_reparent_with_audit_token(mach_port,
req.r_req_id,
op | NAMESPACE_HANDLER_NSPACE_EVENT,
path, dest_path, atoken);
zfree(ZV_NAMEI, dest_path);
} else if (vp_vtype == VDIR) {
char *tmpname = NULL;
/*
* If the caller provided a lookup_name *and* a name length,
* then we assume the lookup_name is not NUL-terminated.
* Allocate a temporary buffer in this case to provide
* a NUL-terminated path name to the IPC call.
*/
if (lookup_name != NULL && namelen != 0) {
if (namelen >= PATH_MAX) {
error = EINVAL;
goto out_req_remove;
}
tmpname = zalloc(ZV_NAMEI);
strlcpy(tmpname, lookup_name, namelen + 1);
lookup_name = tmpname;
} else if (lookup_name != NULL) {
/*
* If the caller provided a lookup_name with a
* zero name length, then we assume it's NUL-
* terminated. Verify it has a valid length.
*/
if (strlen(lookup_name) >= PATH_MAX) {
error = EINVAL;
goto out_req_remove;
}
}
/* (See above.) */
kern_ret = send_vfs_resolve_dir_with_audit_token(mach_port,
req.r_req_id,
op | NAMESPACE_HANDLER_NSPACE_EVENT,
lookup_name == NULL ? "" : lookup_name, path, atoken);
if (tmpname != NULL) {
zfree(ZV_NAMEI, tmpname);
/*
* Poison lookup_name rather than reference
* freed memory.
*/
lookup_name = NULL;
}
} else {
/* (See above.) */
kern_ret = send_vfs_resolve_file_with_audit_token(mach_port,
req.r_req_id,
op | NAMESPACE_HANDLER_NSPACE_EVENT,
offset, size, path, atoken);
}
if (kern_ret != KERN_SUCCESS) {
/*
* Also treat this like being unable to access the backing
* store server.
*/
os_log_error(OS_LOG_DEFAULT, "NSPACE resolve failure: %d",
kern_ret);
error = ETIMEDOUT;
goto out_req_remove;
}
/*
* Give back the memory we allocated earlier while we wait; we
* no longer need it.
*/
kfree_data(path, path_alloc_len);
path = NULL;
/*
* Request has been submitted to the resolver. Now (interruptibly)
* wait for completion. Upon requrn, the request will have been
* removed from the lookup table.
*/
error = nspace_resolver_req_wait(&req);
out_release_port:
if (path != NULL) {
kfree_data(path, path_alloc_len);
path = NULL;
}
ipc_port_release_send(mach_port);
out_check_errors:
/*
* The file resolver owns the logic about what error to return
* to the caller. We only need to handle a couple of special
* cases here:
*/
if (error == EJUSTRETURN) {
/*
* The requesting process is allowed to interact with
* dataless objects. Make a couple of sanity-checks
* here to ensure the action makes sense.
*/
switch (op) {
case NAMESPACE_HANDLER_WRITE_OP:
case NAMESPACE_HANDLER_TRUNCATE_OP:
case NAMESPACE_HANDLER_RENAME_OP:
/*
* This handles the case of the resolver itself
* writing data to the file (or throwing it
* away).
*/
error = 0;
break;
case NAMESPACE_HANDLER_READ_OP:
case NAMESPACE_HANDLER_LOOKUP_OP:
/*
* This handles the case of the resolver needing
* to look up inside of a dataless directory while
* it's in the process of materializing it (for
* example, creating files or directories).
*/
error = (vp_vtype == VDIR) ? 0 : EBADF;
break;
default:
error = EBADF;
break;
}
}
return error;
out_req_remove:
nspace_resolver_req_remove(&req);
goto out_release_port;
#else
return ENOTSUP;
#endif /* CONFIG_DATALESS_FILES */
}
/*
* vfs_materialize_file: Materialize a regular file.
*
* Inputs:
* vp The dataless file to be materialized.
*
* op What kind of operation is being performed:
* -> NAMESPACE_HANDLER_READ_OP
* -> NAMESPACE_HANDLER_WRITE_OP
* -> NAMESPACE_HANDLER_LINK_CREATE
* -> NAMESPACE_HANDLER_DELETE_OP
* -> NAMESPACE_HANDLER_TRUNCATE_OP
* -> NAMESPACE_HANDLER_RENAME_OP
*
* offset offset of I/O for READ or WRITE. Ignored for
* other ops.
*
* size size of I/O for READ or WRITE Ignored for
* other ops.
*
* If offset or size are -1 for a READ or WRITE, then the resolver should
* consider the range to be unknown.
*
* Upon successful return, the caller may proceed with the operation.
* N.B. the file may still be "dataless" in this case.
*/
int
vfs_materialize_file(
struct vnode *vp,
uint64_t op,
int64_t offset,
int64_t size)
{
if (vp->v_type != VREG) {
return EFTYPE;
}
return vfs_materialize_item(vp, (uint32_t)op, offset, size, NULL, 0,
NULL);
}
/*
* vfs_materialize_dir:
*
* Inputs:
* vp The dataless directory to be materialized.
*
* op What kind of operation is being performed:
* -> NAMESPACE_HANDLER_READ_OP
* -> NAMESPACE_HANDLER_WRITE_OP
* -> NAMESPACE_HANDLER_DELETE_OP
* -> NAMESPACE_HANDLER_RENAME_OP
* -> NAMESPACE_HANDLER_LOOKUP_OP
*
* lookup_name Name being looked up for a LOOKUP op. Ignored for
* other ops. May or may not be NUL-terminated; see below.
*
* namelen If non-zero, then lookup_name is assumed to not be NUL-
* terminated and namelen is the number of valid bytes in
* lookup_name. If zero, then lookup_name is assumed to be
* NUL-terminated.
*
* Upon successful return, the caller may proceed with the operation.
* N.B. the directory may still be "dataless" in this case.
*/
int
vfs_materialize_dir(
struct vnode *vp,
uint64_t op,
char *lookup_name,
size_t namelen)
{
if (vp->v_type != VDIR) {
return EFTYPE;
}
if (op == NAMESPACE_HANDLER_LOOKUP_OP && lookup_name == NULL) {
return EINVAL;
}
return vfs_materialize_item(vp, (uint32_t)op, 0, 0, lookup_name,
namelen, NULL);
}
/*
* vfs_materialize_reparent:
*
* Inputs:
* vp The dataless file or directory to be materialized.
*
* tdvp The new parent directory for the dataless file.
*
* Upon successful return, the caller may proceed with the operation.
* N.B. the item may still be "dataless" in this case.
*/
int
vfs_materialize_reparent(vnode_t vp, vnode_t tdvp)
{
if (vp->v_type != VDIR && vp->v_type != VREG) {
return EFTYPE;
}
return vfs_materialize_item(vp, NAMESPACE_HANDLER_RENAME_OP,
0, 0, NULL, 0, tdvp);
}
#if 0
static int
build_volfs_path(struct vnode *vp, char *path, int *len)
{
struct vnode_attr va;
int ret;
VATTR_INIT(&va);
VATTR_WANTED(&va, va_fsid);
VATTR_WANTED(&va, va_fileid);
if (vnode_getattr(vp, &va, vfs_context_kernel()) != 0) {
*len = snprintf(path, *len, "/non/existent/path/because/vnode_getattr/failed") + 1;
ret = -1;
} else {
*len = snprintf(path, *len, "/.vol/%d/%lld", (dev_t)va.va_fsid, va.va_fileid) + 1;
ret = 0;
}
return ret;
}
#endif
static unsigned long
fsctl_bogus_command_compat(unsigned long cmd)
{
switch (cmd) {
case IOCBASECMD(FSIOC_SYNC_VOLUME):
return FSIOC_SYNC_VOLUME;
case IOCBASECMD(FSIOC_ROUTEFS_SETROUTEID):
return FSIOC_ROUTEFS_SETROUTEID;
case IOCBASECMD(FSIOC_SET_PACKAGE_EXTS):
return FSIOC_SET_PACKAGE_EXTS;
case IOCBASECMD(FSIOC_SET_FSTYPENAME_OVERRIDE):
return FSIOC_SET_FSTYPENAME_OVERRIDE;
case IOCBASECMD(DISK_CONDITIONER_IOC_GET):
return DISK_CONDITIONER_IOC_GET;
case IOCBASECMD(DISK_CONDITIONER_IOC_SET):
return DISK_CONDITIONER_IOC_SET;
case IOCBASECMD(FSIOC_FIOSEEKHOLE):
return FSIOC_FIOSEEKHOLE;
case IOCBASECMD(FSIOC_FIOSEEKDATA):
return FSIOC_FIOSEEKDATA;
case IOCBASECMD(SPOTLIGHT_IOC_GET_MOUNT_TIME):
return SPOTLIGHT_IOC_GET_MOUNT_TIME;
case IOCBASECMD(SPOTLIGHT_IOC_GET_LAST_MTIME):
return SPOTLIGHT_IOC_GET_LAST_MTIME;
}
return cmd;
}
static int
cas_bsdflags_setattr(vnode_t vp, void *arg, vfs_context_t ctx)
{
return VNOP_IOCTL(vp, FSIOC_CAS_BSDFLAGS, arg, FWRITE, ctx);
}
static int __attribute__((noinline))
handle_sync_volume(vnode_t vp, vnode_t *arg_vp, caddr_t data, vfs_context_t ctx)
{
struct vfs_attr vfa;
mount_t mp = vp->v_mount;
unsigned arg;
int error;
/* record vid of vp so we can drop it below. */
uint32_t vvid = vp->v_id;
/*
* Then grab mount_iterref so that we can release the vnode.
* Without this, a thread may call vnode_iterate_prepare then
* get into a deadlock because we've never released the root vp
*/
error = mount_iterref(mp, 0);
if (error) {
return error;
}
vnode_hold(vp);
vnode_put(vp);
arg = MNT_NOWAIT;
if (*(uint32_t*)data & FSCTL_SYNC_WAIT) {
arg = MNT_WAIT;
}
/*
* If the filessytem supports multiple filesytems in a
* partition (For eg APFS volumes in a container, it knows
* that the waitfor argument to VFS_SYNC are flags.
*/
VFSATTR_INIT(&vfa);
VFSATTR_WANTED(&vfa, f_capabilities);
if ((vfs_getattr(mp, &vfa, vfs_context_current()) == 0) &&
VFSATTR_IS_SUPPORTED(&vfa, f_capabilities) &&
((vfa.f_capabilities.valid[VOL_CAPABILITIES_FORMAT] & VOL_CAP_FMT_SHARED_SPACE)) &&
((vfa.f_capabilities.capabilities[VOL_CAPABILITIES_FORMAT] & VOL_CAP_FMT_SHARED_SPACE))) {
arg |= MNT_VOLUME;
}
/* issue the sync for this volume */
(void)sync_callback(mp, &arg);
/*
* Then release the mount_iterref once we're done syncing; it's not
* needed for the VNOP_IOCTL below
*/
mount_iterdrop(mp);
if (arg & FSCTL_SYNC_FULLSYNC) {
/* re-obtain vnode iocount on the root vp, if possible */
error = vnode_getwithvid(vp, vvid);
if (error == 0) {
error = VNOP_IOCTL(vp, F_FULLFSYNC, (caddr_t)NULL, 0, ctx);
vnode_put(vp);
}
}
vnode_drop(vp);
/* mark the argument VP as having been released */
*arg_vp = NULL;
return error;
}
#if ROUTEFS
static int __attribute__((noinline))
handle_routes(user_addr_t udata)
{
char routepath[MAXPATHLEN];
size_t len = 0;
int error;
if ((error = suser(kauth_cred_get(), &(current_proc()->p_acflag)))) {
return error;
}
bzero(routepath, MAXPATHLEN);
error = copyinstr(udata, &routepath[0], MAXPATHLEN, &len);
if (error) {
return error;
}
error = routefs_kernel_mount(routepath);
return error;
}
#endif
static int __attribute__((noinline))
handle_flags(vnode_t vp, caddr_t data, vfs_context_t ctx)
{
struct fsioc_cas_bsdflags *cas = (struct fsioc_cas_bsdflags *)data;
struct vnode_attr va;
int error;
VATTR_INIT(&va);
VATTR_SET(&va, va_flags, cas->new_flags);
error = chflags0(vp, &va, cas_bsdflags_setattr, cas, ctx);
#if CONFIG_FSE
if (error == 0 && cas->expected_flags == cas->actual_flags && need_fsevent(FSE_STAT_CHANGED, vp)) {
add_fsevent(FSE_STAT_CHANGED, ctx, FSE_ARG_VNODE, vp, FSE_ARG_DONE);
}
#endif
return error;
}
static int __attribute__((noinline))
handle_auth(vnode_t vp, u_long cmd, caddr_t data, u_long options, vfs_context_t ctx)
{
struct mount *mp = NULL;
errno_t rootauth = 0;
mp = vp->v_mount;
/*
* query the underlying FS and see if it reports something
* sane for this vnode. If volume is authenticated via
* chunklist, leave that for the caller to determine.
*/
rootauth = VNOP_IOCTL(vp, cmd, data, (int)options, ctx);
return rootauth;
}
#define SET_PACKAGE_EXTENSION_ENTITLEMENT \
"com.apple.private.kernel.set-package-extensions"
/*
* Make a filesystem-specific control call:
*/
/* ARGSUSED */
static int
fsctl_internal(proc_t p, vnode_t *arg_vp, u_long cmd, user_addr_t udata, u_long options, vfs_context_t ctx)
{
int error = 0;
boolean_t is64bit;
u_int size;
#define STK_PARAMS 128
char stkbuf[STK_PARAMS] = {0};
caddr_t data, memp;
vnode_t vp = *arg_vp;
if (vp->v_type == VCHR || vp->v_type == VBLK) {
return ENOTTY;
}
cmd = fsctl_bogus_command_compat(cmd);
size = IOCPARM_LEN(cmd);
if (size > IOCPARM_MAX) {
return EINVAL;
}
is64bit = proc_is64bit(p);
memp = NULL;
if (size > sizeof(stkbuf)) {
if ((memp = (caddr_t)kalloc_data(size, Z_WAITOK)) == 0) {
return ENOMEM;
}
data = memp;
} else {
data = &stkbuf[0];
};
if (cmd & IOC_IN) {
if (size) {
error = copyin(udata, data, size);
if (error) {
if (memp) {
kfree_data(memp, size);
}
return error;
}
} else {
if (is64bit) {
*(user_addr_t *)data = udata;
} else {
*(uint32_t *)data = (uint32_t)udata;
}
};
} else if ((cmd & IOC_OUT) && size) {
/*
* Zero the buffer so the user always
* gets back something deterministic.
*/
bzero(data, size);
} else if (cmd & IOC_VOID) {
if (is64bit) {
*(user_addr_t *)data = udata;
} else {
*(uint32_t *)data = (uint32_t)udata;
}
}
/* Check to see if it's a generic command */
switch (cmd) {
case FSIOC_SYNC_VOLUME:
error = handle_sync_volume(vp, arg_vp, data, ctx);
break;
case FSIOC_ROUTEFS_SETROUTEID:
#if ROUTEFS
error = handle_routes(udata);
#endif
break;
case FSIOC_SET_PACKAGE_EXTS: {
user_addr_t ext_strings;
uint32_t num_entries;
uint32_t max_width;
if (!IOTaskHasEntitlement(vfs_context_task(ctx),
SET_PACKAGE_EXTENSION_ENTITLEMENT)) {
error = EPERM;
break;
}
if ((is64bit && size != sizeof(user64_package_ext_info))
|| (is64bit == 0 && size != sizeof(user32_package_ext_info))) {
// either you're 64-bit and passed a 64-bit struct or
// you're 32-bit and passed a 32-bit struct. otherwise
// it's not ok.
error = EINVAL;
break;
}
if (is64bit) {
if (sizeof(user64_addr_t) > sizeof(user_addr_t)) {
assert(((user64_package_ext_info *)data)->strings <= UINT32_MAX);
}
ext_strings = (user_addr_t)((user64_package_ext_info *)data)->strings;
num_entries = ((user64_package_ext_info *)data)->num_entries;
max_width = ((user64_package_ext_info *)data)->max_width;
} else {
ext_strings = CAST_USER_ADDR_T(((user32_package_ext_info *)data)->strings);
num_entries = ((user32_package_ext_info *)data)->num_entries;
max_width = ((user32_package_ext_info *)data)->max_width;
}
error = set_package_extensions_table(ext_strings, num_entries, max_width);
}
break;
case FSIOC_SET_FSTYPENAME_OVERRIDE:
{
mount_t mp;
if ((error = suser(kauth_cred_get(), &(current_proc()->p_acflag)))) {
break;
}
if ((mp = vp->v_mount) != NULL) {
mount_lock(mp);
if (data[0] != 0) {
for (int i = 0; i < MFSTYPENAMELEN; i++) {
if (!data[i]) {
goto continue_copy;
}
}
/*
* Getting here means we have a user data
* string which has no NULL termination in
* its first MFSTYPENAMELEN bytes. This is
* bogus, let's avoid strlcpy-ing the read
* data and return an error.
*/
error = EINVAL;
goto unlock;
continue_copy:
vfs_setfstypename_locked(mp, data);
if (vfs_isrdonly(mp) &&
strcmp(data, "mtmfs") == 0) {
mp->mnt_kern_flag |=
MNTK_EXTENDED_SECURITY;
mp->mnt_kern_flag &=
~MNTK_AUTH_OPAQUE;
}
} else if (mp->mnt_kern_flag & MNTK_TYPENAME_OVERRIDE) {
const char *name =
vfs_getfstypenameref_locked(mp, NULL);
if (strcmp(name, "mtmfs") == 0) {
mp->mnt_kern_flag &=
~MNTK_EXTENDED_SECURITY;
}
vfs_setfstypename_locked(mp, NULL);
}
unlock:
mount_unlock(mp);
}
}
break;
case DISK_CONDITIONER_IOC_GET: {
error = disk_conditioner_get_info(vp->v_mount, (disk_conditioner_info *)data);
}
break;
case DISK_CONDITIONER_IOC_SET: {
error = disk_conditioner_set_info(vp->v_mount, (disk_conditioner_info *)data);
}
break;
case FSIOC_CAS_BSDFLAGS:
error = handle_flags(vp, data, ctx);
break;
case FSIOC_FD_ONLY_OPEN_ONCE: {
error = 0;
if (vnode_usecount(vp) > 1) {
vnode_lock_spin(vp);
if (vp->v_lflag & VL_HASSTREAMS) {
if (vnode_isinuse_locked(vp, 1, 1)) {
error = EBUSY;
}
} else if (vnode_usecount(vp) > 1) {
error = EBUSY;
}
vnode_unlock(vp);
}
}
break;
case FSIOC_EVAL_ROOTAUTH:
error = handle_auth(vp, cmd, data, options, ctx);
break;
case FSIOC_TEST_FSE_ACCESS_GRANTED:
error = test_fse_access_granted(vp, (unsigned long)udata, ctx);
break;
#if CONFIG_EXCLAVES
case FSIOC_EXCLAVE_FS_REGISTER:
if (IOTaskHasEntitlement(vfs_context_task(ctx), EXCLAVE_FS_REGISTER_ENTITLEMENT)) {
error = vfs_exclave_fs_register(((fsioc_exclave_fs_register_t *)data)->fs_tag, vp);
} else {
error = EPERM;
}
break;
case FSIOC_EXCLAVE_FS_UNREGISTER:
if (IOTaskHasEntitlement(vfs_context_task(ctx), EXCLAVE_FS_REGISTER_ENTITLEMENT)) {
error = vfs_exclave_fs_unregister(vp);
} else {
error = EPERM;
}
break;
case FSIOC_EXCLAVE_FS_GET_BASE_DIRS: {
exclave_fs_get_base_dirs_t *get_base_dirs = ((exclave_fs_get_base_dirs_t *)data);
exclave_fs_base_dir_t *dirs = NULL;
if (!IOTaskHasEntitlement(vfs_context_task(ctx), EXCLAVE_FS_REGISTER_ENTITLEMENT)) {
error = EPERM;
break;
}
if (get_base_dirs->base_dirs) {
if ((get_base_dirs->count == 0) || (get_base_dirs->count > EXCLAVE_FS_GET_BASE_DIRS_MAX_COUNT)) {
error = EINVAL;
break;
}
dirs = kalloc_type(exclave_fs_base_dir_t, get_base_dirs->count, Z_WAITOK | Z_ZERO);
if (!dirs) {
error = ENOSPC;
break;
}
}
error = vfs_exclave_fs_get_base_dirs(dirs, &get_base_dirs->count);
if (!error && dirs) {
error = copyout(dirs, (user_addr_t)get_base_dirs->base_dirs,
get_base_dirs->count * sizeof(exclave_fs_base_dir_t));
}
if (dirs) {
kfree_type(exclave_fs_base_dir_t, get_base_dirs->count, dirs);
}
}
break;
#endif
default: {
/*
* Other, known commands shouldn't be passed down here.
* (When adding a selector to this list, it may be prudent
* to consider adding it to the list in sys_fcntl_nocancel() as well.)
*/
switch (cmd) {
case F_PUNCHHOLE:
case F_TRIM_ACTIVE_FILE:
case F_RDADVISE:
case F_TRANSCODEKEY:
case F_GETPROTECTIONLEVEL:
case F_GETDEFAULTPROTLEVEL:
case F_MAKECOMPRESSED:
case F_SET_GREEDY_MODE:
case F_SETSTATICCONTENT:
case F_SETIOTYPE:
case F_SETBACKINGSTORE:
case F_GETPATH_MTMINFO:
case APFSIOC_REVERT_TO_SNAPSHOT:
case FSIOC_FIOSEEKHOLE:
case FSIOC_FIOSEEKDATA:
case HFS_GET_BOOT_INFO:
case HFS_SET_BOOT_INFO:
case FIOPINSWAP:
case F_CHKCLEAN:
case F_FULLFSYNC:
case F_BARRIERFSYNC:
case F_FREEZE_FS:
case F_THAW_FS:
case FSIOC_KERNEL_ROOTAUTH:
case FSIOC_GRAFT_FS:
case FSIOC_UNGRAFT_FS:
case FSIOC_AUTH_FS:
error = EINVAL;
goto outdrop;
}
/* Invoke the filesystem-specific code */
error = VNOP_IOCTL(vp, cmd, data, (int)options, ctx);
}
} /* end switch stmt */
/*
* if no errors, copy any data to user. Size was
* already set and checked above.
*/
if (error == 0 && (cmd & IOC_OUT) && size) {
error = copyout(data, udata, size);
}
outdrop:
if (memp) {
kfree_data(memp, size);
}
return error;
}
/* ARGSUSED */
int
fsctl(proc_t p, struct fsctl_args *uap, __unused int32_t *retval)
{
int error;
struct nameidata nd;
uint32_t nameiflags;
vnode_t vp = NULL;
vfs_context_t ctx = vfs_context_current();
AUDIT_ARG(cmd, (int)uap->cmd);
AUDIT_ARG(value32, uap->options);
/* Get the vnode for the file we are getting info on: */
nameiflags = 0;
//
// if we come through fsctl() then the file is by definition not open.
// therefore for the FSIOC_FD_ONLY_OPEN_ONCE selector we return an error
// lest the caller mistakenly thinks the only open is their own (but in
// reality it's someone elses).
//
if (uap->cmd == FSIOC_FD_ONLY_OPEN_ONCE) {
return EINVAL;
}
if ((uap->options & FSOPT_NOFOLLOW) == 0) {
nameiflags |= FOLLOW;
}
if (uap->cmd == FSIOC_FIRMLINK_CTL) {
nameiflags |= (CN_FIRMLINK_NOFOLLOW | NOCACHE);
}
NDINIT(&nd, LOOKUP, OP_FSCTL, nameiflags | AUDITVNPATH1,
UIO_USERSPACE, uap->path, ctx);
if ((error = namei(&nd))) {
goto done;
}
vp = nd.ni_vp;
nameidone(&nd);
#if CONFIG_MACF
error = mac_mount_check_fsctl(ctx, vnode_mount(vp), uap->cmd);
if (error) {
goto done;
}
#endif
error = fsctl_internal(p, &vp, uap->cmd, (user_addr_t)uap->data, uap->options, ctx);
done:
if (vp) {
vnode_put(vp);
}
return error;
}
/* ARGSUSED */
int
ffsctl(proc_t p, struct ffsctl_args *uap, __unused int32_t *retval)
{
int error;
vnode_t vp = NULL;
vfs_context_t ctx = vfs_context_current();
int fd = -1;
AUDIT_ARG(fd, uap->fd);
AUDIT_ARG(cmd, (int)uap->cmd);
AUDIT_ARG(value32, uap->options);
/* Get the vnode for the file we are getting info on: */
if ((error = file_vnode(uap->fd, &vp))) {
return error;
}
fd = uap->fd;
if ((error = vnode_getwithref(vp))) {
file_drop(fd);
return error;
}
#if CONFIG_MACF
if ((error = mac_mount_check_fsctl(ctx, vnode_mount(vp), uap->cmd))) {
file_drop(fd);
vnode_put(vp);
return error;
}
#endif
error = fsctl_internal(p, &vp, uap->cmd, (user_addr_t)uap->data, uap->options, ctx);
file_drop(fd);
/*validate vp; fsctl_internal() can drop iocount and reset vp to NULL*/
if (vp) {
vnode_put(vp);
}
return error;
}
/* end of fsctl system call */
#define FILESEC_ACCESS_ENTITLEMENT \
"com.apple.private.vfs.filesec-access"
static int
xattr_entitlement_check(const char *attrname, vfs_context_t ctx, bool setting)
{
if (strcmp(attrname, KAUTH_FILESEC_XATTR) == 0) {
/*
* get: root and tasks with FILESEC_ACCESS_ENTITLEMENT.
* set: only tasks with FILESEC_ACCESS_ENTITLEMENT.
*/
if ((!setting && vfs_context_issuser(ctx)) ||
IOTaskHasEntitlement(vfs_context_task(ctx),
FILESEC_ACCESS_ENTITLEMENT)) {
return 0;
}
}
return EPERM;
}
/*
* Retrieve the data of an extended attribute.
*/
int
getxattr(proc_t p, struct getxattr_args *uap, user_ssize_t *retval)
{
vnode_t vp;
struct nameidata nd;
char attrname[XATTR_MAXNAMELEN + 1];
vfs_context_t ctx = vfs_context_current();
uio_t auio = NULL;
int spacetype = IS_64BIT_PROCESS(p) ? UIO_USERSPACE64 : UIO_USERSPACE32;
size_t attrsize = 0;
size_t namelen;
u_int32_t nameiflags;
int error;
UIO_STACKBUF(uio_buf, 1);
if (uap->options & (XATTR_NOSECURITY | XATTR_NODEFAULT)) {
return EINVAL;
}
nameiflags = (uap->options & XATTR_NOFOLLOW) ? 0 : FOLLOW;
NDINIT(&nd, LOOKUP, OP_GETXATTR, nameiflags, spacetype, uap->path, ctx);
if ((error = namei(&nd))) {
return error;
}
vp = nd.ni_vp;
nameidone(&nd);
error = copyinstr(uap->attrname, attrname, sizeof(attrname), &namelen);
if (error != 0) {
goto out;
}
if (xattr_protected(attrname) &&
(error = xattr_entitlement_check(attrname, ctx, false)) != 0) {
goto out;
}
/*
* the specific check for 0xffffffff is a hack to preserve
* binaray compatibilty in K64 with applications that discovered
* that passing in a buf pointer and a size of -1 resulted in
* just the size of the indicated extended attribute being returned.
* this isn't part of the documented behavior, but because of the
* original implemtation's check for "uap->size > 0", this behavior
* was allowed. In K32 that check turned into a signed comparison
* even though uap->size is unsigned... in K64, we blow by that
* check because uap->size is unsigned and doesn't get sign smeared
* in the munger for a 32 bit user app. we also need to add a
* check to limit the maximum size of the buffer being passed in...
* unfortunately, the underlying fileystems seem to just malloc
* the requested size even if the actual extended attribute is tiny.
* because that malloc is for kernel wired memory, we have to put a
* sane limit on it.
*
* U32 running on K64 will yield 0x00000000ffffffff for uap->size
* U64 running on K64 will yield -1 (64 bits wide)
* U32/U64 running on K32 will yield -1 (32 bits wide)
*/
if (uap->size == 0xffffffff || uap->size == (size_t)-1) {
goto no_uio;
}
if (uap->value) {
if (uap->size > (size_t)XATTR_MAXSIZE) {
uap->size = XATTR_MAXSIZE;
}
auio = uio_createwithbuffer(1, uap->position, spacetype, UIO_READ,
&uio_buf[0], sizeof(uio_buf));
uio_addiov(auio, uap->value, uap->size);
}
no_uio:
error = vn_getxattr(vp, attrname, auio, &attrsize, uap->options, ctx);
out:
vnode_put(vp);
if (auio) {
*retval = uap->size - uio_resid(auio);
} else {
*retval = (user_ssize_t)attrsize;
}
return error;
}
/*
* Retrieve the data of an extended attribute.
*/
int
fgetxattr(proc_t p, struct fgetxattr_args *uap, user_ssize_t *retval)
{
vnode_t vp;
char attrname[XATTR_MAXNAMELEN + 1];
vfs_context_t ctx = vfs_context_current();
uio_t auio = NULL;
int spacetype = IS_64BIT_PROCESS(p) ? UIO_USERSPACE64 : UIO_USERSPACE32;
size_t attrsize = 0;
size_t namelen;
int error;
UIO_STACKBUF(uio_buf, 1);
if (uap->options & (XATTR_NOFOLLOW | XATTR_NOSECURITY | XATTR_NODEFAULT)) {
return EINVAL;
}
if ((error = file_vnode(uap->fd, &vp))) {
return error;
}
if ((error = vnode_getwithref(vp))) {
file_drop(uap->fd);
return error;
}
error = copyinstr(uap->attrname, attrname, sizeof(attrname), &namelen);
if (error != 0) {
goto out;
}
if (xattr_protected(attrname) &&
(error = xattr_entitlement_check(attrname, ctx, false)) != 0) {
goto out;
}
if (uap->value && uap->size > 0) {
if (uap->size > (size_t)XATTR_MAXSIZE) {
uap->size = XATTR_MAXSIZE;
}
auio = uio_createwithbuffer(1, uap->position, spacetype, UIO_READ,
&uio_buf[0], sizeof(uio_buf));
uio_addiov(auio, uap->value, uap->size);
}
error = vn_getxattr(vp, attrname, auio, &attrsize, uap->options, vfs_context_current());
out:
(void)vnode_put(vp);
file_drop(uap->fd);
if (auio) {
*retval = uap->size - uio_resid(auio);
} else {
*retval = (user_ssize_t)attrsize;
}
return error;
}
/* struct for checkdirs iteration */
struct setxattr_ctx {
struct nameidata nd;
char attrname[XATTR_MAXNAMELEN + 1];
UIO_STACKBUF(uio_buf, 1);
};
/*
* Set the data of an extended attribute.
*/
int
setxattr(proc_t p, struct setxattr_args *uap, int *retval)
{
vnode_t vp;
vfs_context_t ctx = vfs_context_current();
uio_t auio = NULL;
int spacetype = IS_64BIT_PROCESS(p) ? UIO_USERSPACE64 : UIO_USERSPACE32;
size_t namelen;
u_int32_t nameiflags;
int error;
struct setxattr_ctx *sactx;
if (uap->options & (XATTR_NOSECURITY | XATTR_NODEFAULT)) {
return EINVAL;
}
sactx = kalloc_type(struct setxattr_ctx, Z_WAITOK);
if (sactx == NULL) {
return ENOMEM;
}
error = copyinstr(uap->attrname, sactx->attrname, sizeof(sactx->attrname), &namelen);
if (error != 0) {
if (error == EPERM) {
/* if the string won't fit in attrname, copyinstr emits EPERM */
error = ENAMETOOLONG;
}
/* Otherwise return the default error from copyinstr to detect ERANGE, etc */
goto out;
}
if (xattr_protected(sactx->attrname) &&
(error = xattr_entitlement_check(sactx->attrname, ctx, true)) != 0) {
goto out;
}
if (uap->size != 0 && uap->value == 0) {
error = EINVAL;
goto out;
}
if (uap->size > INT_MAX) {
error = E2BIG;
goto out;
}
nameiflags = (uap->options & XATTR_NOFOLLOW) ? 0 : FOLLOW;
#if CONFIG_FILE_LEASES
nameiflags |= WANTPARENT;
#endif
NDINIT(&sactx->nd, LOOKUP, OP_SETXATTR, nameiflags, spacetype, uap->path, ctx);
if ((error = namei(&sactx->nd))) {
goto out;
}
vp = sactx->nd.ni_vp;
#if CONFIG_FILE_LEASES
vnode_breakdirlease(sactx->nd.ni_dvp, false, O_WRONLY);
vnode_put(sactx->nd.ni_dvp);
#endif
nameidone(&sactx->nd);
auio = uio_createwithbuffer(1, uap->position, spacetype, UIO_WRITE,
&sactx->uio_buf[0], sizeof(sactx->uio_buf));
uio_addiov(auio, uap->value, uap->size);
error = vn_setxattr(vp, sactx->attrname, auio, uap->options, ctx);
#if CONFIG_FSE
if (error == 0) {
add_fsevent(FSE_XATTR_MODIFIED, ctx,
FSE_ARG_VNODE, vp,
FSE_ARG_DONE);
}
#endif
vnode_put(vp);
out:
kfree_type(struct setxattr_ctx, sactx);
*retval = 0;
return error;
}
/*
* Set the data of an extended attribute.
*/
int
fsetxattr(proc_t p, struct fsetxattr_args *uap, int *retval)
{
vnode_t vp;
char attrname[XATTR_MAXNAMELEN + 1];
vfs_context_t ctx = vfs_context_current();
uio_t auio = NULL;
int spacetype = IS_64BIT_PROCESS(p) ? UIO_USERSPACE64 : UIO_USERSPACE32;
size_t namelen;
int error;
UIO_STACKBUF(uio_buf, 1);
if (uap->options & (XATTR_NOFOLLOW | XATTR_NOSECURITY | XATTR_NODEFAULT)) {
return EINVAL;
}
error = copyinstr(uap->attrname, attrname, sizeof(attrname), &namelen);
if (error != 0) {
if (error == EPERM) {
/* if the string won't fit in attrname, copyinstr emits EPERM */
return ENAMETOOLONG;
}
/* Otherwise return the default error from copyinstr to detect ERANGE, etc */
return error;
}
if (xattr_protected(attrname) &&
(error = xattr_entitlement_check(attrname, ctx, true)) != 0) {
return error;
}
if (uap->size != 0 && uap->value == 0) {
return EINVAL;
}
if (uap->size > INT_MAX) {
return E2BIG;
}
if ((error = file_vnode(uap->fd, &vp))) {
return error;
}
if ((error = vnode_getwithref(vp))) {
file_drop(uap->fd);
return error;
}
#if CONFIG_FILE_LEASES
vnode_breakdirlease(vp, true, O_WRONLY);
#endif
auio = uio_createwithbuffer(1, uap->position, spacetype, UIO_WRITE,
&uio_buf[0], sizeof(uio_buf));
uio_addiov(auio, uap->value, uap->size);
error = vn_setxattr(vp, attrname, auio, uap->options, vfs_context_current());
#if CONFIG_FSE
if (error == 0) {
add_fsevent(FSE_XATTR_MODIFIED, ctx,
FSE_ARG_VNODE, vp,
FSE_ARG_DONE);
}
#endif
vnode_put(vp);
file_drop(uap->fd);
*retval = 0;
return error;
}
/*
* Remove an extended attribute.
* XXX Code duplication here.
*/
int
removexattr(proc_t p, struct removexattr_args *uap, int *retval)
{
vnode_t vp;
struct nameidata nd;
char attrname[XATTR_MAXNAMELEN + 1];
int spacetype = IS_64BIT_PROCESS(p) ? UIO_USERSPACE64 : UIO_USERSPACE32;
vfs_context_t ctx = vfs_context_current();
size_t namelen;
u_int32_t nameiflags;
int error;
if (uap->options & (XATTR_NOSECURITY | XATTR_NODEFAULT)) {
return EINVAL;
}
error = copyinstr(uap->attrname, attrname, sizeof(attrname), &namelen);
if (error != 0) {
return error;
}
if (xattr_protected(attrname)) {
return EPERM;
}
nameiflags = (uap->options & XATTR_NOFOLLOW) ? 0 : FOLLOW;
#if CONFIG_FILE_LEASES
nameiflags |= WANTPARENT;
#endif
NDINIT(&nd, LOOKUP, OP_REMOVEXATTR, nameiflags, spacetype, uap->path, ctx);
if ((error = namei(&nd))) {
return error;
}
vp = nd.ni_vp;
#if CONFIG_FILE_LEASES
vnode_breakdirlease(nd.ni_dvp, false, O_WRONLY);
vnode_put(nd.ni_dvp);
#endif
nameidone(&nd);
error = vn_removexattr(vp, attrname, uap->options, ctx);
#if CONFIG_FSE
if (error == 0) {
add_fsevent(FSE_XATTR_REMOVED, ctx,
FSE_ARG_VNODE, vp,
FSE_ARG_DONE);
}
#endif
vnode_put(vp);
*retval = 0;
return error;
}
/*
* Remove an extended attribute.
* XXX Code duplication here.
*/
int
fremovexattr(__unused proc_t p, struct fremovexattr_args *uap, int *retval)
{
vnode_t vp;
char attrname[XATTR_MAXNAMELEN + 1];
size_t namelen;
int error;
#if CONFIG_FSE
vfs_context_t ctx = vfs_context_current();
#endif
if (uap->options & (XATTR_NOFOLLOW | XATTR_NOSECURITY | XATTR_NODEFAULT)) {
return EINVAL;
}
error = copyinstr(uap->attrname, attrname, sizeof(attrname), &namelen);
if (error != 0) {
return error;
}
if (xattr_protected(attrname)) {
return EPERM;
}
if ((error = file_vnode(uap->fd, &vp))) {
return error;
}
if ((error = vnode_getwithref(vp))) {
file_drop(uap->fd);
return error;
}
#if CONFIG_FILE_LEASES
vnode_breakdirlease(vp, true, O_WRONLY);
#endif
error = vn_removexattr(vp, attrname, uap->options, vfs_context_current());
#if CONFIG_FSE
if (error == 0) {
add_fsevent(FSE_XATTR_REMOVED, ctx,
FSE_ARG_VNODE, vp,
FSE_ARG_DONE);
}
#endif
vnode_put(vp);
file_drop(uap->fd);
*retval = 0;
return error;
}
/*
* Retrieve the list of extended attribute names.
* XXX Code duplication here.
*/
int
listxattr(proc_t p, struct listxattr_args *uap, user_ssize_t *retval)
{
vnode_t vp;
struct nameidata nd;
vfs_context_t ctx = vfs_context_current();
uio_t auio = NULL;
int spacetype = IS_64BIT_PROCESS(p) ? UIO_USERSPACE64 : UIO_USERSPACE32;
size_t attrsize = 0;
u_int32_t nameiflags;
int error;
UIO_STACKBUF(uio_buf, 1);
if (uap->options & (XATTR_NOSECURITY | XATTR_NODEFAULT)) {
return EINVAL;
}
nameiflags = (uap->options & XATTR_NOFOLLOW) ? 0 : FOLLOW;
NDINIT(&nd, LOOKUP, OP_LISTXATTR, nameiflags, spacetype, uap->path, ctx);
if ((error = namei(&nd))) {
return error;
}
vp = nd.ni_vp;
nameidone(&nd);
if (uap->namebuf != 0 && uap->bufsize > 0) {
auio = uio_createwithbuffer(1, 0, spacetype, UIO_READ,
&uio_buf[0], sizeof(uio_buf));
uio_addiov(auio, uap->namebuf, uap->bufsize);
}
error = vn_listxattr(vp, auio, &attrsize, uap->options, ctx);
vnode_put(vp);
if (auio) {
*retval = (user_ssize_t)uap->bufsize - uio_resid(auio);
} else {
*retval = (user_ssize_t)attrsize;
}
return error;
}
/*
* Retrieve the list of extended attribute names.
* XXX Code duplication here.
*/
int
flistxattr(proc_t p, struct flistxattr_args *uap, user_ssize_t *retval)
{
vnode_t vp;
uio_t auio = NULL;
int spacetype = proc_is64bit(p) ? UIO_USERSPACE64 : UIO_USERSPACE32;
size_t attrsize = 0;
int error;
UIO_STACKBUF(uio_buf, 1);
if (uap->options & (XATTR_NOFOLLOW | XATTR_NOSECURITY | XATTR_NODEFAULT)) {
return EINVAL;
}
if ((error = file_vnode(uap->fd, &vp))) {
return error;
}
if ((error = vnode_getwithref(vp))) {
file_drop(uap->fd);
return error;
}
if (uap->namebuf != 0 && uap->bufsize > 0) {
auio = uio_createwithbuffer(1, 0, spacetype,
UIO_READ, &uio_buf[0], sizeof(uio_buf));
uio_addiov(auio, uap->namebuf, uap->bufsize);
}
error = vn_listxattr(vp, auio, &attrsize, uap->options, vfs_context_current());
vnode_put(vp);
file_drop(uap->fd);
if (auio) {
*retval = (user_ssize_t)uap->bufsize - uio_resid(auio);
} else {
*retval = (user_ssize_t)attrsize;
}
return error;
}
int
fsgetpath_internal(vfs_context_t ctx, int volfs_id, uint64_t objid,
vm_size_t bufsize, caddr_t buf, uint32_t options, int *pathlen)
{
int error;
struct mount *mp = NULL;
vnode_t vp;
int length;
int bpflags;
/* maximum number of times to retry build_path */
unsigned int retries = 0x10;
if (bufsize > FSGETPATH_MAXBUFLEN) {
return EINVAL;
}
if (buf == NULL) {
return ENOMEM;
}
retry:
if ((mp = mount_lookupby_volfsid(volfs_id, 1)) == NULL) {
error = ENOTSUP; /* unexpected failure */
return ENOTSUP;
}
#if CONFIG_UNION_MOUNTS
unionget:
#endif /* CONFIG_UNION_MOUNTS */
if (objid == 2) {
struct vfs_attr vfsattr;
int use_vfs_root = TRUE;
VFSATTR_INIT(&vfsattr);
VFSATTR_WANTED(&vfsattr, f_capabilities);
if (!(options & FSOPT_ISREALFSID) &&
vfs_getattr(mp, &vfsattr, vfs_context_kernel()) == 0 &&
VFSATTR_IS_SUPPORTED(&vfsattr, f_capabilities)) {
if ((vfsattr.f_capabilities.capabilities[VOL_CAPABILITIES_FORMAT] & VOL_CAP_FMT_VOL_GROUPS) &&
(vfsattr.f_capabilities.valid[VOL_CAPABILITIES_FORMAT] & VOL_CAP_FMT_VOL_GROUPS)) {
use_vfs_root = FALSE;
}
}
if (use_vfs_root) {
error = VFS_ROOT(mp, &vp, ctx);
} else {
error = VFS_VGET(mp, objid, &vp, ctx);
}
} else {
error = VFS_VGET(mp, (ino64_t)objid, &vp, ctx);
}
#if CONFIG_UNION_MOUNTS
if (error == ENOENT && (mp->mnt_flag & MNT_UNION)) {
/*
* If the fileid isn't found and we're in a union
* mount volume, then see if the fileid is in the
* mounted-on volume.
*/
struct mount *tmp = mp;
mp = vnode_mount(tmp->mnt_vnodecovered);
vfs_unbusy(tmp);
if (vfs_busy(mp, LK_NOWAIT) == 0) {
goto unionget;
}
} else {
vfs_unbusy(mp);
}
#else
vfs_unbusy(mp);
#endif /* CONFIG_UNION_MOUNTS */
if (error) {
return error;
}
#if CONFIG_MACF
error = mac_vnode_check_fsgetpath(ctx, vp);
if (error) {
vnode_put(vp);
return error;
}
#endif
/* Obtain the absolute path to this vnode. */
bpflags = vfs_context_suser(ctx) ? BUILDPATH_CHECKACCESS : 0;
if (options & FSOPT_NOFIRMLINKPATH) {
bpflags |= BUILDPATH_NO_FIRMLINK;
}
bpflags |= BUILDPATH_CHECK_MOVED;
error = build_path(vp, buf, (int)bufsize, &length, bpflags, ctx);
vnode_put(vp);
if (error) {
/* there was a race building the path, try a few more times */
if (error == EAGAIN) {
--retries;
if (retries > 0) {
goto retry;
}
error = ENOENT;
}
goto out;
}
AUDIT_ARG(text, buf);
if (kdebug_debugid_enabled(VFS_LOOKUP) && length > 0) {
unsigned long path_words[NUMPARMS];
size_t path_len = sizeof(path_words);
if ((size_t)length < path_len) {
memcpy((char *)path_words, buf, length);
memset((char *)path_words + length, 0, path_len - length);
path_len = length;
} else {
memcpy((char *)path_words, buf + (length - path_len), path_len);
}
kdebug_vfs_lookup(path_words, (int)path_len, vp,
KDBG_VFS_LOOKUP_FLAG_LOOKUP);
}
*pathlen = length; /* may be superseded by error */
out:
return error;
}
/*
* Obtain the full pathname of a file system object by id.
*/
static int
fsgetpath_extended(user_addr_t buf, user_size_t bufsize, user_addr_t user_fsid, uint64_t objid,
uint32_t options, user_ssize_t *retval)
{
vfs_context_t ctx = vfs_context_current();
fsid_t fsid;
char *realpath;
int length;
int error;
if (options & ~(FSOPT_NOFIRMLINKPATH | FSOPT_ISREALFSID)) {
return EINVAL;
}
if ((error = copyin(user_fsid, (caddr_t)&fsid, sizeof(fsid)))) {
return error;
}
AUDIT_ARG(value32, fsid.val[0]);
AUDIT_ARG(value64, objid);
/* Restrict output buffer size for now. */
if (bufsize > FSGETPATH_MAXBUFLEN || bufsize <= 0) {
return EINVAL;
}
realpath = kalloc_data(bufsize, Z_WAITOK | Z_ZERO);
if (realpath == NULL) {
return ENOMEM;
}
error = fsgetpath_internal(ctx, fsid.val[0], objid, bufsize, realpath,
options, &length);
if (error) {
goto out;
}
error = copyout((caddr_t)realpath, buf, length);
*retval = (user_ssize_t)length; /* may be superseded by error */
out:
kfree_data(realpath, bufsize);
return error;
}
int
fsgetpath(__unused proc_t p, struct fsgetpath_args *uap, user_ssize_t *retval)
{
return fsgetpath_extended(uap->buf, uap->bufsize, uap->fsid, uap->objid,
0, retval);
}
int
fsgetpath_ext(__unused proc_t p, struct fsgetpath_ext_args *uap, user_ssize_t *retval)
{
return fsgetpath_extended(uap->buf, uap->bufsize, uap->fsid, uap->objid,
uap->options, retval);
}
/*
* Common routine to handle various flavors of statfs data heading out
* to user space.
*
* Returns: 0 Success
* EFAULT
*/
static int
munge_statfs(struct mount *mp, struct vfsstatfs *sfsp,
user_addr_t bufp, int *sizep, boolean_t is_64_bit,
boolean_t partial_copy)
{
int error;
int my_size, copy_size;
if (is_64_bit) {
struct user64_statfs sfs;
my_size = copy_size = sizeof(sfs);
bzero(&sfs, my_size);
sfs.f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
sfs.f_type = (short)mp->mnt_vtable->vfc_typenum;
sfs.f_reserved1 = (short)sfsp->f_fssubtype;
sfs.f_bsize = (user64_long_t)sfsp->f_bsize;
sfs.f_iosize = (user64_long_t)sfsp->f_iosize;
sfs.f_blocks = (user64_long_t)sfsp->f_blocks;
sfs.f_bfree = (user64_long_t)sfsp->f_bfree;
sfs.f_bavail = (user64_long_t)sfsp->f_bavail;
sfs.f_files = (user64_long_t)sfsp->f_files;
sfs.f_ffree = (user64_long_t)sfsp->f_ffree;
sfs.f_fsid = sfsp->f_fsid;
sfs.f_owner = sfsp->f_owner;
vfs_getfstypename(mp, sfs.f_fstypename, MFSNAMELEN);
strlcpy(&sfs.f_mntonname[0], &sfsp->f_mntonname[0], MNAMELEN);
strlcpy(&sfs.f_mntfromname[0], &sfsp->f_mntfromname[0], MNAMELEN);
if (partial_copy) {
copy_size -= (sizeof(sfs.f_reserved3) + sizeof(sfs.f_reserved4));
}
error = copyout((caddr_t)&sfs, bufp, copy_size);
} else {
struct user32_statfs sfs;
my_size = copy_size = sizeof(sfs);
bzero(&sfs, my_size);
sfs.f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
sfs.f_type = (short)mp->mnt_vtable->vfc_typenum;
sfs.f_reserved1 = (short)sfsp->f_fssubtype;
/*
* It's possible for there to be more than 2^^31 blocks in the filesystem, so we
* have to fudge the numbers here in that case. We inflate the blocksize in order
* to reflect the filesystem size as best we can.
*/
if ((sfsp->f_blocks > INT_MAX)
/* Hack for 4061702 . I think the real fix is for Carbon to
* look for some volume capability and not depend on hidden
* semantics agreed between a FS and carbon.
* f_blocks, f_bfree, and f_bavail set to -1 is the trigger
* for Carbon to set bNoVolumeSizes volume attribute.
* Without this the webdavfs files cannot be copied onto
* disk as they look huge. This change should not affect
* XSAN as they should not setting these to -1..
*/
&& (sfsp->f_blocks != 0xffffffffffffffffULL)
&& (sfsp->f_bfree != 0xffffffffffffffffULL)
&& (sfsp->f_bavail != 0xffffffffffffffffULL)) {
int shift;
/*
* Work out how far we have to shift the block count down to make it fit.
* Note that it's possible to have to shift so far that the resulting
* blocksize would be unreportably large. At that point, we will clip
* any values that don't fit.
*
* For safety's sake, we also ensure that f_iosize is never reported as
* being smaller than f_bsize.
*/
for (shift = 0; shift < 32; shift++) {
if ((sfsp->f_blocks >> shift) <= INT_MAX) {
break;
}
if ((sfsp->f_bsize << (shift + 1)) > INT_MAX) {
break;
}
}
#define __SHIFT_OR_CLIP(x, s) ((((x) >> (s)) > INT_MAX) ? INT_MAX : ((x) >> (s)))
sfs.f_blocks = (user32_long_t)__SHIFT_OR_CLIP(sfsp->f_blocks, shift);
sfs.f_bfree = (user32_long_t)__SHIFT_OR_CLIP(sfsp->f_bfree, shift);
sfs.f_bavail = (user32_long_t)__SHIFT_OR_CLIP(sfsp->f_bavail, shift);
#undef __SHIFT_OR_CLIP
sfs.f_bsize = (user32_long_t)(sfsp->f_bsize << shift);
sfs.f_iosize = (int)lmax(sfsp->f_iosize, sfsp->f_bsize);
} else {
/* filesystem is small enough to be reported honestly */
sfs.f_bsize = (user32_long_t)sfsp->f_bsize;
sfs.f_iosize = (user32_long_t)sfsp->f_iosize;
sfs.f_blocks = (user32_long_t)sfsp->f_blocks;
sfs.f_bfree = (user32_long_t)sfsp->f_bfree;
sfs.f_bavail = (user32_long_t)sfsp->f_bavail;
}
sfs.f_files = (user32_long_t)sfsp->f_files;
sfs.f_ffree = (user32_long_t)sfsp->f_ffree;
sfs.f_fsid = sfsp->f_fsid;
sfs.f_owner = sfsp->f_owner;
vfs_getfstypename(mp, sfs.f_fstypename, MFSNAMELEN);
strlcpy(&sfs.f_mntonname[0], &sfsp->f_mntonname[0], MNAMELEN);
strlcpy(&sfs.f_mntfromname[0], &sfsp->f_mntfromname[0], MNAMELEN);
if (partial_copy) {
copy_size -= (sizeof(sfs.f_reserved3) + sizeof(sfs.f_reserved4));
}
error = copyout((caddr_t)&sfs, bufp, copy_size);
}
if (sizep != NULL) {
*sizep = my_size;
}
return error;
}
/*
* copy stat structure into user_stat structure.
*/
void
munge_user64_stat(struct stat *sbp, struct user64_stat *usbp)
{
bzero(usbp, sizeof(*usbp));
usbp->st_dev = sbp->st_dev;
usbp->st_ino = sbp->st_ino;
usbp->st_mode = sbp->st_mode;
usbp->st_nlink = sbp->st_nlink;
usbp->st_uid = sbp->st_uid;
usbp->st_gid = sbp->st_gid;
usbp->st_rdev = sbp->st_rdev;
#ifndef _POSIX_C_SOURCE
usbp->st_atimespec.tv_sec = sbp->st_atimespec.tv_sec;
usbp->st_atimespec.tv_nsec = sbp->st_atimespec.tv_nsec;
usbp->st_mtimespec.tv_sec = sbp->st_mtimespec.tv_sec;
usbp->st_mtimespec.tv_nsec = sbp->st_mtimespec.tv_nsec;
usbp->st_ctimespec.tv_sec = sbp->st_ctimespec.tv_sec;
usbp->st_ctimespec.tv_nsec = sbp->st_ctimespec.tv_nsec;
#else
usbp->st_atime = sbp->st_atime;
usbp->st_atimensec = sbp->st_atimensec;
usbp->st_mtime = sbp->st_mtime;
usbp->st_mtimensec = sbp->st_mtimensec;
usbp->st_ctime = sbp->st_ctime;
usbp->st_ctimensec = sbp->st_ctimensec;
#endif
usbp->st_size = sbp->st_size;
usbp->st_blocks = sbp->st_blocks;
usbp->st_blksize = sbp->st_blksize;
usbp->st_flags = sbp->st_flags;
usbp->st_gen = sbp->st_gen;
usbp->st_lspare = sbp->st_lspare;
usbp->st_qspare[0] = sbp->st_qspare[0];
usbp->st_qspare[1] = sbp->st_qspare[1];
}
void
munge_user32_stat(struct stat *sbp, struct user32_stat *usbp)
{
bzero(usbp, sizeof(*usbp));
usbp->st_dev = sbp->st_dev;
usbp->st_ino = sbp->st_ino;
usbp->st_mode = sbp->st_mode;
usbp->st_nlink = sbp->st_nlink;
usbp->st_uid = sbp->st_uid;
usbp->st_gid = sbp->st_gid;
usbp->st_rdev = sbp->st_rdev;
#ifndef _POSIX_C_SOURCE
usbp->st_atimespec.tv_sec = (user32_time_t)sbp->st_atimespec.tv_sec;
usbp->st_atimespec.tv_nsec = (user32_long_t)sbp->st_atimespec.tv_nsec;
usbp->st_mtimespec.tv_sec = (user32_time_t)sbp->st_mtimespec.tv_sec;
usbp->st_mtimespec.tv_nsec = (user32_long_t)sbp->st_mtimespec.tv_nsec;
usbp->st_ctimespec.tv_sec = (user32_time_t)sbp->st_ctimespec.tv_sec;
usbp->st_ctimespec.tv_nsec = (user32_long_t)sbp->st_ctimespec.tv_nsec;
#else
usbp->st_atime = sbp->st_atime;
usbp->st_atimensec = sbp->st_atimensec;
usbp->st_mtime = sbp->st_mtime;
usbp->st_mtimensec = sbp->st_mtimensec;
usbp->st_ctime = sbp->st_ctime;
usbp->st_ctimensec = sbp->st_ctimensec;
#endif
usbp->st_size = sbp->st_size;
usbp->st_blocks = sbp->st_blocks;
usbp->st_blksize = sbp->st_blksize;
usbp->st_flags = sbp->st_flags;
usbp->st_gen = sbp->st_gen;
usbp->st_lspare = sbp->st_lspare;
usbp->st_qspare[0] = sbp->st_qspare[0];
usbp->st_qspare[1] = sbp->st_qspare[1];
}
/*
* copy stat64 structure into user_stat64 structure.
*/
void
munge_user64_stat64(struct stat64 *sbp, struct user64_stat64 *usbp)
{
bzero(usbp, sizeof(*usbp));
usbp->st_dev = sbp->st_dev;
usbp->st_ino = sbp->st_ino;
usbp->st_mode = sbp->st_mode;
usbp->st_nlink = sbp->st_nlink;
usbp->st_uid = sbp->st_uid;
usbp->st_gid = sbp->st_gid;
usbp->st_rdev = sbp->st_rdev;
#ifndef _POSIX_C_SOURCE
usbp->st_atimespec.tv_sec = sbp->st_atimespec.tv_sec;
usbp->st_atimespec.tv_nsec = sbp->st_atimespec.tv_nsec;
usbp->st_mtimespec.tv_sec = sbp->st_mtimespec.tv_sec;
usbp->st_mtimespec.tv_nsec = sbp->st_mtimespec.tv_nsec;
usbp->st_ctimespec.tv_sec = sbp->st_ctimespec.tv_sec;
usbp->st_ctimespec.tv_nsec = sbp->st_ctimespec.tv_nsec;
usbp->st_birthtimespec.tv_sec = sbp->st_birthtimespec.tv_sec;
usbp->st_birthtimespec.tv_nsec = sbp->st_birthtimespec.tv_nsec;
#else
usbp->st_atime = sbp->st_atime;
usbp->st_atimensec = sbp->st_atimensec;
usbp->st_mtime = sbp->st_mtime;
usbp->st_mtimensec = sbp->st_mtimensec;
usbp->st_ctime = sbp->st_ctime;
usbp->st_ctimensec = sbp->st_ctimensec;
usbp->st_birthtime = sbp->st_birthtime;
usbp->st_birthtimensec = sbp->st_birthtimensec;
#endif
usbp->st_size = sbp->st_size;
usbp->st_blocks = sbp->st_blocks;
usbp->st_blksize = sbp->st_blksize;
usbp->st_flags = sbp->st_flags;
usbp->st_gen = sbp->st_gen;
usbp->st_lspare = sbp->st_lspare;
usbp->st_qspare[0] = sbp->st_qspare[0];
usbp->st_qspare[1] = sbp->st_qspare[1];
}
void
munge_user32_stat64(struct stat64 *sbp, struct user32_stat64 *usbp)
{
bzero(usbp, sizeof(*usbp));
usbp->st_dev = sbp->st_dev;
usbp->st_ino = sbp->st_ino;
usbp->st_mode = sbp->st_mode;
usbp->st_nlink = sbp->st_nlink;
usbp->st_uid = sbp->st_uid;
usbp->st_gid = sbp->st_gid;
usbp->st_rdev = sbp->st_rdev;
#ifndef _POSIX_C_SOURCE
usbp->st_atimespec.tv_sec = (user32_time_t)sbp->st_atimespec.tv_sec;
usbp->st_atimespec.tv_nsec = (user32_long_t)sbp->st_atimespec.tv_nsec;
usbp->st_mtimespec.tv_sec = (user32_time_t)sbp->st_mtimespec.tv_sec;
usbp->st_mtimespec.tv_nsec = (user32_long_t)sbp->st_mtimespec.tv_nsec;
usbp->st_ctimespec.tv_sec = (user32_time_t)sbp->st_ctimespec.tv_sec;
usbp->st_ctimespec.tv_nsec = (user32_long_t)sbp->st_ctimespec.tv_nsec;
usbp->st_birthtimespec.tv_sec = (user32_time_t)sbp->st_birthtimespec.tv_sec;
usbp->st_birthtimespec.tv_nsec = (user32_long_t)sbp->st_birthtimespec.tv_nsec;
#else
usbp->st_atime = sbp->st_atime;
usbp->st_atimensec = sbp->st_atimensec;
usbp->st_mtime = sbp->st_mtime;
usbp->st_mtimensec = sbp->st_mtimensec;
usbp->st_ctime = sbp->st_ctime;
usbp->st_ctimensec = sbp->st_ctimensec;
usbp->st_birthtime = sbp->st_birthtime;
usbp->st_birthtimensec = sbp->st_birthtimensec;
#endif
usbp->st_size = sbp->st_size;
usbp->st_blocks = sbp->st_blocks;
usbp->st_blksize = sbp->st_blksize;
usbp->st_flags = sbp->st_flags;
usbp->st_gen = sbp->st_gen;
usbp->st_lspare = sbp->st_lspare;
usbp->st_qspare[0] = sbp->st_qspare[0];
usbp->st_qspare[1] = sbp->st_qspare[1];
}
/*
* Purge buffer cache for simulating cold starts
*/
static int
vnode_purge_callback(struct vnode *vp, __unused void *cargs)
{
ubc_msync(vp, (off_t)0, ubc_getsize(vp), NULL /* off_t *resid_off */, UBC_PUSHALL | UBC_INVALIDATE);
return VNODE_RETURNED;
}
static int
vfs_purge_callback(mount_t mp, __unused void * arg)
{
vnode_iterate(mp, VNODE_WAIT | VNODE_ITERATE_ALL, vnode_purge_callback, NULL);
return VFS_RETURNED;
}
static TUNABLE_WRITEABLE(boolean_t, vfs_purge_vm_pagers, "vfs_purge_vm_pagers", TRUE);
SYSCTL_INT(_vfs, OID_AUTO, purge_vm_pagers, CTLFLAG_RW | CTLFLAG_LOCKED, &vfs_purge_vm_pagers, 0, "VFS purge also purges file-backed VM pagers");
int
vfs_purge(__unused struct proc *p, __unused struct vfs_purge_args *uap, __unused int32_t *retval)
{
if (!kauth_cred_issuser(kauth_cred_get())) {
return EPERM;
}
vfs_iterate(0 /* flags */, vfs_purge_callback, NULL);
/* also flush any VM pagers backed by files */
if (vfs_purge_vm_pagers) {
vm_purge_filebacked_pagers();
}
return 0;
}
/*
* gets the vnode associated with the (unnamed) snapshot directory
* for a Filesystem. The snapshot directory vnode is returned with
* an iocount on it.
*/
int
vnode_get_snapdir(vnode_t rvp, vnode_t *sdvpp, vfs_context_t ctx)
{
return VFS_VGET_SNAPDIR(vnode_mount(rvp), sdvpp, ctx);
}
/*
* Get the snapshot vnode.
*
* If successful, the call returns with an iocount on *rvpp ,*sdvpp and
* needs nameidone() on ndp.
*
* If the snapshot vnode exists it is returned in ndp->ni_vp.
*
* If it returns with an error, *rvpp, *sdvpp are NULL and nameidone() is
* not needed.
*/
static int
vnode_get_snapshot(int dirfd, vnode_t *rvpp, vnode_t *sdvpp,
user_addr_t name, struct nameidata *ndp, int32_t op,
#if !CONFIG_TRIGGERS
__unused
#endif
enum path_operation pathop,
vfs_context_t ctx)
{
int error, i;
caddr_t name_buf;
size_t name_len;
struct vfs_attr vfa;
*sdvpp = NULLVP;
*rvpp = NULLVP;
error = vnode_getfromfd(ctx, dirfd, rvpp);
if (error) {
return error;
}
if (!vnode_isvroot(*rvpp)) {
error = EINVAL;
goto out;
}
/* Make sure the filesystem supports snapshots */
VFSATTR_INIT(&vfa);
VFSATTR_WANTED(&vfa, f_capabilities);
if ((vfs_getattr(vnode_mount(*rvpp), &vfa, ctx) != 0) ||
!VFSATTR_IS_SUPPORTED(&vfa, f_capabilities) ||
!((vfa.f_capabilities.valid[VOL_CAPABILITIES_INTERFACES] &
VOL_CAP_INT_SNAPSHOT)) ||
!((vfa.f_capabilities.capabilities[VOL_CAPABILITIES_INTERFACES] &
VOL_CAP_INT_SNAPSHOT))) {
error = ENOTSUP;
goto out;
}
error = vnode_get_snapdir(*rvpp, sdvpp, ctx);
if (error) {
goto out;
}
name_buf = zalloc_flags(ZV_NAMEI, Z_WAITOK);
error = copyinstr(name, name_buf, MAXPATHLEN, &name_len);
if (error) {
goto out1;
}
/*
* Some sanity checks- name can't be empty, "." or ".." or have slashes.
* (the length returned by copyinstr includes the terminating NUL)
*/
if ((name_len == 1) || (name_len == 2 && name_buf[0] == '.') ||
(name_len == 3 && name_buf[0] == '.' && name_buf[1] == '.')) {
error = EINVAL;
goto out1;
}
for (i = 0; i < (int)name_len && name_buf[i] != '/'; i++) {
;
}
if (i < (int)name_len) {
error = EINVAL;
goto out1;
}
#if CONFIG_MACF
if (op == CREATE) {
error = mac_mount_check_snapshot_create(ctx, vnode_mount(*rvpp),
name_buf);
} else if (op == DELETE) {
error = mac_mount_check_snapshot_delete(ctx, vnode_mount(*rvpp),
name_buf);
}
if (error) {
goto out1;
}
#endif
/* Check if the snapshot already exists ... */
NDINIT(ndp, op, pathop, USEDVP | NOCACHE | AUDITVNPATH1,
UIO_SYSSPACE, CAST_USER_ADDR_T(name_buf), ctx);
ndp->ni_dvp = *sdvpp;
error = namei(ndp);
out1:
zfree(ZV_NAMEI, name_buf);
out:
if (error) {
if (*sdvpp) {
vnode_put(*sdvpp);
*sdvpp = NULLVP;
}
if (*rvpp) {
vnode_put(*rvpp);
*rvpp = NULLVP;
}
}
return error;
}
/*
* create a filesystem snapshot (for supporting filesystems)
*
* A much simplified version of openat(dirfd, name, O_CREAT | O_EXCL)
* We get to the (unnamed) snapshot directory vnode and create the vnode
* for the snapshot in it.
*
* Restrictions:
*
* a) Passed in name for snapshot cannot have slashes.
* b) name can't be "." or ".."
*
* Since this requires superuser privileges, vnode_authorize calls are not
* made.
*/
static int __attribute__((noinline))
snapshot_create(int dirfd, user_addr_t name, __unused uint32_t flags,
vfs_context_t ctx)
{
vnode_t rvp, snapdvp;
int error;
struct nameidata *ndp;
ndp = kalloc_type(struct nameidata, Z_WAITOK);
error = vnode_get_snapshot(dirfd, &rvp, &snapdvp, name, ndp, CREATE,
OP_LINK, ctx);
if (error) {
goto out;
}
if (ndp->ni_vp) {
vnode_put(ndp->ni_vp);
error = EEXIST;
} else {
struct vnode_attr *vap;
vnode_t vp = NULLVP;
vap = kalloc_type(struct vnode_attr, Z_WAITOK);
VATTR_INIT(vap);
VATTR_SET(vap, va_type, VREG);
VATTR_SET(vap, va_mode, 0);
error = vn_create(snapdvp, &vp, ndp, vap,
VN_CREATE_NOAUTH | VN_CREATE_NOINHERIT, 0, NULL, ctx);
if (!error && vp) {
vnode_put(vp);
}
kfree_type(struct vnode_attr, vap);
}
nameidone(ndp);
vnode_put(snapdvp);
vnode_put(rvp);
out:
kfree_type(struct nameidata, ndp);
return error;
}
/*
* Delete a Filesystem snapshot
*
* get the vnode for the unnamed snapshot directory and the snapshot and
* delete the snapshot.
*/
static int __attribute__((noinline))
snapshot_delete(int dirfd, user_addr_t name, __unused uint32_t flags,
vfs_context_t ctx)
{
vnode_t rvp, snapdvp;
int error;
struct nameidata *ndp;
ndp = kalloc_type(struct nameidata, Z_WAITOK);
error = vnode_get_snapshot(dirfd, &rvp, &snapdvp, name, ndp, DELETE,
OP_UNLINK, ctx);
if (error) {
goto out;
}
error = VNOP_REMOVE(snapdvp, ndp->ni_vp, &ndp->ni_cnd,
VNODE_REMOVE_SKIP_NAMESPACE_EVENT, ctx);
vnode_put(ndp->ni_vp);
nameidone(ndp);
vnode_put(snapdvp);
vnode_put(rvp);
out:
kfree_type(struct nameidata, ndp);
return error;
}
/*
* Revert a filesystem to a snapshot
*
* Marks the filesystem to revert to the given snapshot on next mount.
*/
static int __attribute__((noinline))
snapshot_revert(int dirfd, user_addr_t name, __unused uint32_t flags,
vfs_context_t ctx)
{
int error;
vnode_t rvp;
mount_t mp;
struct fs_snapshot_revert_args revert_data;
struct componentname cnp;
caddr_t name_buf;
size_t name_len;
error = vnode_getfromfd(ctx, dirfd, &rvp);
if (error) {
return error;
}
mp = vnode_mount(rvp);
name_buf = zalloc_flags(ZV_NAMEI, Z_WAITOK);
error = copyinstr(name, name_buf, MAXPATHLEN, &name_len);
if (error) {
zfree(ZV_NAMEI, name_buf);
vnode_put(rvp);
return error;
}
#if CONFIG_MACF
error = mac_mount_check_snapshot_revert(ctx, mp, name_buf);
if (error) {
zfree(ZV_NAMEI, name_buf);
vnode_put(rvp);
return error;
}
#endif
/*
* Grab mount_iterref so that we can release the vnode,
* since VFSIOC_REVERT_SNAPSHOT could conceivably cause a sync.
*/
error = mount_iterref(mp, 0);
vnode_put(rvp);
if (error) {
zfree(ZV_NAMEI, name_buf);
return error;
}
memset(&cnp, 0, sizeof(cnp));
cnp.cn_pnbuf = (char *)name_buf;
cnp.cn_nameiop = LOOKUP;
cnp.cn_flags = ISLASTCN | HASBUF;
cnp.cn_pnlen = MAXPATHLEN;
cnp.cn_nameptr = cnp.cn_pnbuf;
cnp.cn_namelen = (int)name_len;
revert_data.sr_cnp = &cnp;
error = VFS_IOCTL(mp, VFSIOC_REVERT_SNAPSHOT, (caddr_t)&revert_data, 0, ctx);
mount_iterdrop(mp);
zfree(ZV_NAMEI, name_buf);
if (error) {
/* If there was any error, try again using VNOP_IOCTL */
vnode_t snapdvp;
struct nameidata namend;
error = vnode_get_snapshot(dirfd, &rvp, &snapdvp, name, &namend, LOOKUP,
OP_LOOKUP, ctx);
if (error) {
return error;
}
error = VNOP_IOCTL(namend.ni_vp, APFSIOC_REVERT_TO_SNAPSHOT, (caddr_t) NULL,
0, ctx);
vnode_put(namend.ni_vp);
nameidone(&namend);
vnode_put(snapdvp);
vnode_put(rvp);
}
return error;
}
/*
* rename a Filesystem snapshot
*
* get the vnode for the unnamed snapshot directory and the snapshot and
* rename the snapshot. This is a very specialised (and simple) case of
* rename(2) (which has to deal with a lot more complications). It differs
* slightly from rename(2) in that EEXIST is returned if the new name exists.
*/
static int __attribute__((noinline))
snapshot_rename(int dirfd, user_addr_t old, user_addr_t new,
__unused uint32_t flags, vfs_context_t ctx)
{
vnode_t rvp, snapdvp;
int error, i;
caddr_t newname_buf;
size_t name_len;
vnode_t fvp;
struct nameidata *fromnd, *tond;
/* carving out a chunk for structs that are too big to be on stack. */
struct {
struct nameidata from_node;
struct nameidata to_node;
} * __rename_data;
__rename_data = kalloc_type(typeof(*__rename_data), Z_WAITOK);
fromnd = &__rename_data->from_node;
tond = &__rename_data->to_node;
error = vnode_get_snapshot(dirfd, &rvp, &snapdvp, old, fromnd, DELETE,
OP_UNLINK, ctx);
if (error) {
goto out;
}
fvp = fromnd->ni_vp;
newname_buf = zalloc_flags(ZV_NAMEI, Z_WAITOK);
error = copyinstr(new, newname_buf, MAXPATHLEN, &name_len);
if (error) {
goto out1;
}
/*
* Some sanity checks- new name can't be empty, "." or ".." or have
* slashes.
* (the length returned by copyinstr includes the terminating NUL)
*
* The FS rename VNOP is suppossed to handle this but we'll pick it
* off here itself.
*/
if ((name_len == 1) || (name_len == 2 && newname_buf[0] == '.') ||
(name_len == 3 && newname_buf[0] == '.' && newname_buf[1] == '.')) {
error = EINVAL;
goto out1;
}
for (i = 0; i < (int)name_len && newname_buf[i] != '/'; i++) {
;
}
if (i < (int)name_len) {
error = EINVAL;
goto out1;
}
#if CONFIG_MACF
error = mac_mount_check_snapshot_create(ctx, vnode_mount(rvp),
newname_buf);
if (error) {
goto out1;
}
#endif
NDINIT(tond, RENAME, OP_RENAME, USEDVP | NOCACHE | AUDITVNPATH2,
UIO_SYSSPACE, CAST_USER_ADDR_T(newname_buf), ctx);
tond->ni_dvp = snapdvp;
error = namei(tond);
if (error) {
goto out2;
} else if (tond->ni_vp) {
/*
* snapshot rename behaves differently than rename(2) - if the
* new name exists, EEXIST is returned.
*/
vnode_put(tond->ni_vp);
error = EEXIST;
goto out2;
}
error = VNOP_RENAME(snapdvp, fvp, &fromnd->ni_cnd, snapdvp, NULLVP,
&tond->ni_cnd, ctx);
out2:
nameidone(tond);
out1:
zfree(ZV_NAMEI, newname_buf);
vnode_put(fvp);
vnode_put(snapdvp);
vnode_put(rvp);
nameidone(fromnd);
out:
kfree_type(typeof(*__rename_data), __rename_data);
return error;
}
/*
* Mount a Filesystem snapshot
*
* get the vnode for the unnamed snapshot directory and the snapshot and
* mount the snapshot.
*/
static int __attribute__((noinline))
snapshot_mount(int dirfd, user_addr_t name, user_addr_t directory,
__unused user_addr_t mnt_data, __unused uint32_t flags, vfs_context_t ctx)
{
mount_t mp;
vnode_t rvp, snapdvp, snapvp, vp, pvp;
struct fs_snapshot_mount_args smnt_data;
int error;
struct nameidata *snapndp, *dirndp;
/* carving out a chunk for structs that are too big to be on stack. */
struct {
struct nameidata snapnd;
struct nameidata dirnd;
} * __snapshot_mount_data;
__snapshot_mount_data = kalloc_type(typeof(*__snapshot_mount_data), Z_WAITOK);
snapndp = &__snapshot_mount_data->snapnd;
dirndp = &__snapshot_mount_data->dirnd;
error = vnode_get_snapshot(dirfd, &rvp, &snapdvp, name, snapndp, LOOKUP,
OP_LOOKUP, ctx);
if (error) {
goto out;
}
snapvp = snapndp->ni_vp;
if (!vnode_mount(rvp) || (vnode_mount(rvp) == dead_mountp)) {
error = EIO;
goto out1;
}
/* Get the vnode to be covered */
NDINIT(dirndp, LOOKUP, OP_MOUNT, FOLLOW | AUDITVNPATH1 | WANTPARENT,
UIO_USERSPACE, directory, ctx);
error = namei(dirndp);
if (error) {
goto out1;
}
vp = dirndp->ni_vp;
pvp = dirndp->ni_dvp;
mp = vnode_mount(rvp);
if ((vp->v_flag & VROOT) && (vp->v_mount->mnt_flag & MNT_ROOTFS)) {
error = EINVAL;
goto out2;
}
#if CONFIG_MACF
error = mac_mount_check_snapshot_mount(ctx, rvp, vp, &dirndp->ni_cnd, snapndp->ni_cnd.cn_nameptr,
mp->mnt_vfsstat.f_fstypename);
if (error) {
goto out2;
}
#endif
smnt_data.sm_mp = mp;
smnt_data.sm_cnp = &snapndp->ni_cnd;
error = mount_common(mp->mnt_vfsstat.f_fstypename, pvp, vp,
&dirndp->ni_cnd, CAST_USER_ADDR_T(&smnt_data), flags & (MNT_DONTBROWSE | MNT_IGNORE_OWNERSHIP),
KERNEL_MOUNT_SNAPSHOT, NULL, ctx);
out2:
vnode_put(vp);
vnode_put(pvp);
nameidone(dirndp);
out1:
vnode_put(snapvp);
vnode_put(snapdvp);
vnode_put(rvp);
nameidone(snapndp);
out:
kfree_type(typeof(*__snapshot_mount_data), __snapshot_mount_data);
return error;
}
/*
* Root from a snapshot of the filesystem
*
* Marks the filesystem to root from the given snapshot on next boot.
*/
static int __attribute__((noinline))
snapshot_root(int dirfd, user_addr_t name, __unused uint32_t flags,
vfs_context_t ctx)
{
int error;
vnode_t rvp;
mount_t mp;
struct fs_snapshot_root_args root_data;
struct componentname cnp;
caddr_t name_buf;
size_t name_len;
error = vnode_getfromfd(ctx, dirfd, &rvp);
if (error) {
return error;
}
mp = vnode_mount(rvp);
name_buf = zalloc_flags(ZV_NAMEI, Z_WAITOK);
error = copyinstr(name, name_buf, MAXPATHLEN, &name_len);
if (error) {
zfree(ZV_NAMEI, name_buf);
vnode_put(rvp);
return error;
}
// XXX MAC checks ?
/*
* Grab mount_iterref so that we can release the vnode,
* since VFSIOC_ROOT_SNAPSHOT could conceivably cause a sync.
*/
error = mount_iterref(mp, 0);
vnode_put(rvp);
if (error) {
zfree(ZV_NAMEI, name_buf);
return error;
}
memset(&cnp, 0, sizeof(cnp));
cnp.cn_pnbuf = (char *)name_buf;
cnp.cn_nameiop = LOOKUP;
cnp.cn_flags = ISLASTCN | HASBUF;
cnp.cn_pnlen = MAXPATHLEN;
cnp.cn_nameptr = cnp.cn_pnbuf;
cnp.cn_namelen = (int)name_len;
root_data.sr_cnp = &cnp;
error = VFS_IOCTL(mp, VFSIOC_ROOT_SNAPSHOT, (caddr_t)&root_data, 0, ctx);
mount_iterdrop(mp);
zfree(ZV_NAMEI, name_buf);
return error;
}
static boolean_t
vfs_context_can_snapshot(vfs_context_t ctx)
{
static const char * const snapshot_entitlements[] = {
"com.apple.private.vfs.snapshot",
"com.apple.developer.vfs.snapshot",
"com.apple.private.apfs.arv.limited.snapshot",
};
static const size_t nentitlements =
sizeof(snapshot_entitlements) / sizeof(snapshot_entitlements[0]);
size_t i;
task_t task = vfs_context_task(ctx);
for (i = 0; i < nentitlements; i++) {
if (IOTaskHasEntitlement(task, snapshot_entitlements[i])) {
return TRUE;
}
}
return FALSE;
}
/*
* FS snapshot operations dispatcher
*/
int
fs_snapshot(__unused proc_t p, struct fs_snapshot_args *uap,
__unused int32_t *retval)
{
int error;
vfs_context_t ctx = vfs_context_current();
AUDIT_ARG(fd, uap->dirfd);
AUDIT_ARG(value32, uap->op);
if (!vfs_context_can_snapshot(ctx)) {
return EPERM;
}
/*
* Enforce user authorization for snapshot modification operations,
* or if trying to root from snapshot.
*/
if (uap->op != SNAPSHOT_OP_MOUNT) {
vnode_t dvp = NULLVP;
vnode_t devvp = NULLVP;
mount_t mp;
error = vnode_getfromfd(ctx, uap->dirfd, &dvp);
if (error) {
return error;
}
mp = vnode_mount(dvp);
devvp = mp->mnt_devvp;
/* get an iocount on devvp */
if (devvp == NULLVP) {
error = vnode_lookup(mp->mnt_vfsstat.f_mntfromname, 0, &devvp, ctx);
/* for mounts which arent block devices */
if (error == ENOENT) {
error = ENXIO;
}
} else {
error = vnode_getwithref(devvp);
}
if (error) {
vnode_put(dvp);
return error;
}
if ((vfs_context_issuser(ctx) == 0) &&
(vnode_authorize(devvp, NULL, KAUTH_VNODE_WRITE_DATA, ctx) != 0) &&
(!IOTaskHasEntitlement(vfs_context_task(ctx), "com.apple.private.vfs.snapshot.user"))) {
error = EPERM;
}
vnode_put(dvp);
vnode_put(devvp);
if (error) {
return error;
}
}
switch (uap->op) {
case SNAPSHOT_OP_CREATE:
error = snapshot_create(uap->dirfd, uap->name1, uap->flags, ctx);
break;
case SNAPSHOT_OP_DELETE:
error = snapshot_delete(uap->dirfd, uap->name1, uap->flags, ctx);
break;
case SNAPSHOT_OP_RENAME:
error = snapshot_rename(uap->dirfd, uap->name1, uap->name2,
uap->flags, ctx);
break;
case SNAPSHOT_OP_MOUNT:
error = snapshot_mount(uap->dirfd, uap->name1, uap->name2,
uap->data, uap->flags, ctx);
break;
case SNAPSHOT_OP_REVERT:
error = snapshot_revert(uap->dirfd, uap->name1, uap->flags, ctx);
break;
#if CONFIG_MNT_ROOTSNAP
case SNAPSHOT_OP_ROOT:
error = snapshot_root(uap->dirfd, uap->name1, uap->flags, ctx);
break;
#endif /* CONFIG_MNT_ROOTSNAP */
default:
error = ENOSYS;
}
return error;
}
Reference in a new issue Copy permalink