/*
Unix SMB/CIFS implementation.
Wrap disk only vfs functions to sidestep dodgy compilers.
Copyright (C) Tim Potter 1998
Copyright (C) Jeremy Allison 2007
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 3 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program. If not, see .
*/
#include "includes.h"
#undef DBGC_CLASS
#define DBGC_CLASS DBGC_VFS
/* Check for NULL pointer parameters in vfswrap_* functions */
/* We don't want to have NULL function pointers lying around. Someone
is sure to try and execute them. These stubs are used to prevent
this possibility. */
static int vfswrap_connect(vfs_handle_struct *handle, const char *service, const char *user)
{
return 0; /* Return >= 0 for success */
}
static void vfswrap_disconnect(vfs_handle_struct *handle)
{
}
/* Disk operations */
static uint64_t vfswrap_disk_free(vfs_handle_struct *handle, const char *path, bool small_query, uint64_t *bsize,
uint64_t *dfree, uint64_t *dsize)
{
uint64_t result;
result = sys_disk_free(handle->conn, path, small_query, bsize, dfree, dsize);
return result;
}
static int vfswrap_get_quota(struct vfs_handle_struct *handle, enum SMB_QUOTA_TYPE qtype, unid_t id, SMB_DISK_QUOTA *qt)
{
#ifdef HAVE_SYS_QUOTAS
int result;
START_PROFILE(syscall_get_quota);
result = sys_get_quota(handle->conn->connectpath, qtype, id, qt);
END_PROFILE(syscall_get_quota);
return result;
#else
errno = ENOSYS;
return -1;
#endif
}
static int vfswrap_set_quota(struct vfs_handle_struct *handle, enum SMB_QUOTA_TYPE qtype, unid_t id, SMB_DISK_QUOTA *qt)
{
#ifdef HAVE_SYS_QUOTAS
int result;
START_PROFILE(syscall_set_quota);
result = sys_set_quota(handle->conn->connectpath, qtype, id, qt);
END_PROFILE(syscall_set_quota);
return result;
#else
errno = ENOSYS;
return -1;
#endif
}
static int vfswrap_get_shadow_copy_data(struct vfs_handle_struct *handle, struct files_struct *fsp, SHADOW_COPY_DATA *shadow_copy_data, bool labels)
{
errno = ENOSYS;
return -1; /* Not implemented. */
}
static int vfswrap_statvfs(struct vfs_handle_struct *handle, const char *path, vfs_statvfs_struct *statbuf)
{
return sys_statvfs(path, statbuf);
}
static uint32_t vfswrap_fs_capabilities(struct vfs_handle_struct *handle)
{
#if defined(DARWINOS)
struct vfs_statvfs_struct statbuf;
ZERO_STRUCT(statbuf);
sys_statvfs(handle->conn->connectpath, &statbuf);
return statbuf.FsCapabilities;
#endif
return FILE_CASE_SENSITIVE_SEARCH | FILE_CASE_PRESERVED_NAMES;
}
/* Directory operations */
static SMB_STRUCT_DIR *vfswrap_opendir(vfs_handle_struct *handle, const char *fname, const char *mask, uint32 attr)
{
SMB_STRUCT_DIR *result;
START_PROFILE(syscall_opendir);
result = sys_opendir(fname);
END_PROFILE(syscall_opendir);
return result;
}
static SMB_STRUCT_DIRENT *vfswrap_readdir(vfs_handle_struct *handle, SMB_STRUCT_DIR *dirp)
{
SMB_STRUCT_DIRENT *result;
START_PROFILE(syscall_readdir);
result = sys_readdir(dirp);
END_PROFILE(syscall_readdir);
return result;
}
static void vfswrap_seekdir(vfs_handle_struct *handle, SMB_STRUCT_DIR *dirp, long offset)
{
START_PROFILE(syscall_seekdir);
sys_seekdir(dirp, offset);
END_PROFILE(syscall_seekdir);
}
static long vfswrap_telldir(vfs_handle_struct *handle, SMB_STRUCT_DIR *dirp)
{
long result;
START_PROFILE(syscall_telldir);
result = sys_telldir(dirp);
END_PROFILE(syscall_telldir);
return result;
}
static void vfswrap_rewinddir(vfs_handle_struct *handle, SMB_STRUCT_DIR *dirp)
{
START_PROFILE(syscall_rewinddir);
sys_rewinddir(dirp);
END_PROFILE(syscall_rewinddir);
}
static int vfswrap_mkdir(vfs_handle_struct *handle, const char *path, mode_t mode)
{
int result;
bool has_dacl = False;
START_PROFILE(syscall_mkdir);
if (lp_inherit_acls(SNUM(handle->conn)) && (has_dacl = directory_has_default_acl(handle->conn, parent_dirname(path))))
mode = 0777;
result = mkdir(path, mode);
if (result == 0 && !has_dacl) {
/*
* We need to do this as the default behavior of POSIX ACLs
* is to set the mask to be the requested group permission
* bits, not the group permission bits to be the requested
* group permission bits. This is not what we want, as it will
* mess up any inherited ACL bits that were set. JRA.
*/
int saved_errno = errno; /* We may get ENOSYS */
if ((SMB_VFS_CHMOD_ACL(handle->conn, path, mode) == -1) && (errno == ENOSYS))
errno = saved_errno;
}
END_PROFILE(syscall_mkdir);
return result;
}
static int vfswrap_rmdir(vfs_handle_struct *handle, const char *path)
{
int result;
START_PROFILE(syscall_rmdir);
result = rmdir(path);
END_PROFILE(syscall_rmdir);
return result;
}
static int vfswrap_closedir(vfs_handle_struct *handle, SMB_STRUCT_DIR *dirp)
{
int result;
START_PROFILE(syscall_closedir);
result = sys_closedir(dirp);
END_PROFILE(syscall_closedir);
return result;
}
/* File operations */
static int vfswrap_open(vfs_handle_struct *handle, const char *fname,
files_struct *fsp, int flags, mode_t mode)
{
int result;
START_PROFILE(syscall_open);
result = sys_open(fname, flags, mode);
END_PROFILE(syscall_open);
return result;
}
static NTSTATUS vfswrap_create_file(vfs_handle_struct *handle,
struct smb_request *req,
uint16_t root_dir_fid,
const char *fname,
uint32_t create_file_flags,
uint32_t access_mask,
uint32_t share_access,
uint32_t create_disposition,
uint32_t create_options,
uint32_t file_attributes,
uint32_t oplock_request,
uint64_t allocation_size,
struct security_descriptor *sd,
struct ea_list *ea_list,
files_struct **result,
int *pinfo,
SMB_STRUCT_STAT *psbuf)
{
return create_file_default(handle->conn, req, root_dir_fid, fname,
create_file_flags, access_mask, share_access,
create_disposition, create_options,
file_attributes, oplock_request,
allocation_size, sd, ea_list, result, pinfo,
psbuf);
}
static int vfswrap_close(vfs_handle_struct *handle, files_struct *fsp)
{
int result;
START_PROFILE(syscall_close);
result = fd_close_posix(fsp);
END_PROFILE(syscall_close);
return result;
}
static ssize_t vfswrap_read(vfs_handle_struct *handle, files_struct *fsp, void *data, size_t n)
{
ssize_t result;
START_PROFILE_BYTES(syscall_read, n);
result = sys_read(fsp->fh->fd, data, n);
END_PROFILE(syscall_read);
return result;
}
static ssize_t vfswrap_pread(vfs_handle_struct *handle, files_struct *fsp, void *data,
size_t n, SMB_OFF_T offset)
{
ssize_t result;
#if defined(HAVE_PREAD) || defined(HAVE_PREAD64)
START_PROFILE_BYTES(syscall_pread, n);
result = sys_pread(fsp->fh->fd, data, n, offset);
END_PROFILE(syscall_pread);
if (result == -1 && errno == ESPIPE) {
/* Maintain the fiction that pipes can be seeked (sought?) on. */
result = SMB_VFS_READ(fsp, data, n);
fsp->fh->pos = 0;
}
#else /* HAVE_PREAD */
SMB_OFF_T curr;
int lerrno;
curr = SMB_VFS_LSEEK(fsp, 0, SEEK_CUR);
if (curr == -1 && errno == ESPIPE) {
/* Maintain the fiction that pipes can be seeked (sought?) on. */
result = SMB_VFS_READ(fsp, data, n);
fsp->fh->pos = 0;
return result;
}
if (SMB_VFS_LSEEK(fsp, offset, SEEK_SET) == -1) {
return -1;
}
errno = 0;
result = SMB_VFS_READ(fsp, data, n);
lerrno = errno;
SMB_VFS_LSEEK(fsp, curr, SEEK_SET);
errno = lerrno;
#endif /* HAVE_PREAD */
return result;
}
static ssize_t vfswrap_write(vfs_handle_struct *handle, files_struct *fsp, const void *data, size_t n)
{
ssize_t result;
START_PROFILE_BYTES(syscall_write, n);
result = sys_write(fsp->fh->fd, data, n);
END_PROFILE(syscall_write);
return result;
}
static ssize_t vfswrap_pwrite(vfs_handle_struct *handle, files_struct *fsp, const void *data,
size_t n, SMB_OFF_T offset)
{
ssize_t result;
#if defined(HAVE_PWRITE) || defined(HAVE_PRWITE64)
START_PROFILE_BYTES(syscall_pwrite, n);
result = sys_pwrite(fsp->fh->fd, data, n, offset);
END_PROFILE(syscall_pwrite);
if (result == -1 && errno == ESPIPE) {
/* Maintain the fiction that pipes can be sought on. */
result = SMB_VFS_WRITE(fsp, data, n);
}
#else /* HAVE_PWRITE */
SMB_OFF_T curr;
int lerrno;
curr = SMB_VFS_LSEEK(fsp, 0, SEEK_CUR);
if (curr == -1) {
return -1;
}
if (SMB_VFS_LSEEK(fsp, offset, SEEK_SET) == -1) {
return -1;
}
result = SMB_VFS_WRITE(fsp, data, n);
lerrno = errno;
SMB_VFS_LSEEK(fsp, curr, SEEK_SET);
errno = lerrno;
#endif /* HAVE_PWRITE */
return result;
}
static SMB_OFF_T vfswrap_lseek(vfs_handle_struct *handle, files_struct *fsp, SMB_OFF_T offset, int whence)
{
SMB_OFF_T result = 0;
START_PROFILE(syscall_lseek);
/* Cope with 'stat' file opens. */
if (fsp->fh->fd != -1)
result = sys_lseek(fsp->fh->fd, offset, whence);
/*
* We want to maintain the fiction that we can seek
* on a fifo for file system purposes. This allows
* people to set up UNIX fifo's that feed data to Windows
* applications. JRA.
*/
if((result == -1) && (errno == ESPIPE)) {
result = 0;
errno = 0;
}
END_PROFILE(syscall_lseek);
return result;
}
static ssize_t vfswrap_sendfile(vfs_handle_struct *handle, int tofd, files_struct *fromfsp, const DATA_BLOB *hdr,
SMB_OFF_T offset, size_t n)
{
ssize_t result;
START_PROFILE_BYTES(syscall_sendfile, n);
result = sys_sendfile(tofd, fromfsp->fh->fd, hdr, offset, n);
END_PROFILE(syscall_sendfile);
return result;
}
static ssize_t vfswrap_recvfile(vfs_handle_struct *handle,
int fromfd,
files_struct *tofsp,
SMB_OFF_T offset,
size_t n)
{
ssize_t result;
START_PROFILE_BYTES(syscall_recvfile, n);
result = sys_recvfile(fromfd, tofsp->fh->fd, offset, n);
END_PROFILE(syscall_recvfile);
return result;
}
/*********************************************************
For rename across filesystems Patch from Warren Birnbaum
**********************************************************/
static int copy_reg(const char *source, const char *dest)
{
SMB_STRUCT_STAT source_stats;
int saved_errno;
int ifd = -1;
int ofd = -1;
if (sys_lstat (source, &source_stats) == -1)
return -1;
if (!S_ISREG (source_stats.st_mode))
return -1;
if((ifd = sys_open (source, O_RDONLY, 0)) < 0)
return -1;
if (unlink (dest) && errno != ENOENT)
return -1;
#ifdef O_NOFOLLOW
if((ofd = sys_open (dest, O_WRONLY | O_CREAT | O_TRUNC | O_NOFOLLOW, 0600)) < 0 )
#else
if((ofd = sys_open (dest, O_WRONLY | O_CREAT | O_TRUNC , 0600)) < 0 )
#endif
goto err;
if (transfer_file(ifd, ofd, (size_t)-1) == -1)
goto err;
/*
* Try to preserve ownership. For non-root it might fail, but that's ok.
* But root probably wants to know, e.g. if NFS disallows it.
*/
#ifdef HAVE_FCHOWN
if ((fchown(ofd, source_stats.st_uid, source_stats.st_gid) == -1) && (errno != EPERM))
#else
if ((chown(dest, source_stats.st_uid, source_stats.st_gid) == -1) && (errno != EPERM))
#endif
goto err;
/*
* fchown turns off set[ug]id bits for non-root,
* so do the chmod last.
*/
#if defined(HAVE_FCHMOD)
if (fchmod (ofd, source_stats.st_mode & 07777))
#else
if (chmod (dest, source_stats.st_mode & 07777))
#endif
goto err;
if (close (ifd) == -1)
goto err;
if (close (ofd) == -1)
return -1;
/* Try to copy the old file's modtime and access time. */
{
struct utimbuf tv;
tv.actime = source_stats.st_atime;
tv.modtime = source_stats.st_mtime;
utime(dest, &tv);
}
if (unlink (source) == -1)
return -1;
return 0;
err:
saved_errno = errno;
if (ifd != -1)
close(ifd);
if (ofd != -1)
close(ofd);
errno = saved_errno;
return -1;
}
static int vfswrap_rename(vfs_handle_struct *handle, const char *oldname, const char *newname)
{
int result;
START_PROFILE(syscall_rename);
result = rename(oldname, newname);
if ((result == -1) && (errno == EXDEV)) {
/* Rename across filesystems needed. */
result = copy_reg(oldname, newname);
}
END_PROFILE(syscall_rename);
return result;
}
static int vfswrap_fsync(vfs_handle_struct *handle, files_struct *fsp)
{
#ifdef HAVE_FSYNC
int result;
START_PROFILE(syscall_fsync);
result = fsync(fsp->fh->fd);
END_PROFILE(syscall_fsync);
return result;
#else
return 0;
#endif
}
static int vfswrap_stat(vfs_handle_struct *handle, const char *fname, SMB_STRUCT_STAT *sbuf)
{
int result;
START_PROFILE(syscall_stat);
result = sys_stat(fname, sbuf);
END_PROFILE(syscall_stat);
return result;
}
static int vfswrap_fstat(vfs_handle_struct *handle, files_struct *fsp, SMB_STRUCT_STAT *sbuf)
{
int result;
START_PROFILE(syscall_fstat);
result = sys_fstat(fsp->fh->fd, sbuf);
END_PROFILE(syscall_fstat);
return result;
}
int vfswrap_lstat(vfs_handle_struct *handle, const char *path, SMB_STRUCT_STAT *sbuf)
{
int result;
START_PROFILE(syscall_lstat);
result = sys_lstat(path, sbuf);
END_PROFILE(syscall_lstat);
return result;
}
static int vfswrap_unlink(vfs_handle_struct *handle, const char *path)
{
int result;
START_PROFILE(syscall_unlink);
result = unlink(path);
END_PROFILE(syscall_unlink);
return result;
}
static int vfswrap_chmod(vfs_handle_struct *handle, const char *path, mode_t mode)
{
int result;
START_PROFILE(syscall_chmod);
/*
* We need to do this due to the fact that the default POSIX ACL
* chmod modifies the ACL *mask* for the group owner, not the
* group owner bits directly. JRA.
*/
{
int saved_errno = errno; /* We might get ENOSYS */
if ((result = SMB_VFS_CHMOD_ACL(handle->conn, path, mode)) == 0) {
END_PROFILE(syscall_chmod);
return result;
}
/* Error - return the old errno. */
errno = saved_errno;
}
result = chmod(path, mode);
END_PROFILE(syscall_chmod);
return result;
}
static int vfswrap_fchmod(vfs_handle_struct *handle, files_struct *fsp, mode_t mode)
{
int result;
START_PROFILE(syscall_fchmod);
/*
* We need to do this due to the fact that the default POSIX ACL
* chmod modifies the ACL *mask* for the group owner, not the
* group owner bits directly. JRA.
*/
{
int saved_errno = errno; /* We might get ENOSYS */
if ((result = SMB_VFS_FCHMOD_ACL(fsp, mode)) == 0) {
END_PROFILE(syscall_fchmod);
return result;
}
/* Error - return the old errno. */
errno = saved_errno;
}
#if defined(HAVE_FCHMOD)
result = fchmod(fsp->fh->fd, mode);
#else
result = -1;
errno = ENOSYS;
#endif
END_PROFILE(syscall_fchmod);
return result;
}
static int vfswrap_chown(vfs_handle_struct *handle, const char *path, uid_t uid, gid_t gid)
{
int result;
START_PROFILE(syscall_chown);
result = chown(path, uid, gid);
END_PROFILE(syscall_chown);
return result;
}
static int vfswrap_fchown(vfs_handle_struct *handle, files_struct *fsp, uid_t uid, gid_t gid)
{
#ifdef HAVE_FCHOWN
int result;
START_PROFILE(syscall_fchown);
result = fchown(fsp->fh->fd, uid, gid);
END_PROFILE(syscall_fchown);
return result;
#else
errno = ENOSYS;
return -1;
#endif
}
static int vfswrap_lchown(vfs_handle_struct *handle, const char *path, uid_t uid, gid_t gid)
{
int result;
START_PROFILE(syscall_lchown);
result = lchown(path, uid, gid);
END_PROFILE(syscall_lchown);
return result;
}
static int vfswrap_chdir(vfs_handle_struct *handle, const char *path)
{
int result;
START_PROFILE(syscall_chdir);
result = chdir(path);
END_PROFILE(syscall_chdir);
return result;
}
static char *vfswrap_getwd(vfs_handle_struct *handle, char *path)
{
char *result;
START_PROFILE(syscall_getwd);
result = sys_getwd(path);
END_PROFILE(syscall_getwd);
return result;
}
/*********************************************************************
nsec timestamp resolution call. Convert down to whatever the underlying
system will support.
**********************************************************************/
static int vfswrap_ntimes(vfs_handle_struct *handle, const char *path, const struct timespec ts[2])
{
int result;
START_PROFILE(syscall_ntimes);
#if defined(HAVE_UTIMES)
if (ts != NULL) {
struct timeval tv[2];
tv[0] = convert_timespec_to_timeval(ts[0]);
tv[1] = convert_timespec_to_timeval(ts[1]);
result = utimes(path, tv);
} else {
result = utimes(path, NULL);
}
#elif defined(HAVE_UTIME)
if (ts != NULL) {
struct utimbuf times;
times.actime = convert_timespec_to_time_t(ts[0]);
times.modtime = convert_timespec_to_time_t(ts[1]);
result = utime(path, times);
} else {
result = utime(path, NULL);
}
#else
errno = ENOSYS;
result = -1;
#endif
END_PROFILE(syscall_ntimes);
return result;
}
/*********************************************************************
A version of ftruncate that will write the space on disk if strict
allocate is set.
**********************************************************************/
static int strict_allocate_ftruncate(vfs_handle_struct *handle, files_struct *fsp, SMB_OFF_T len)
{
SMB_STRUCT_STAT st;
SMB_OFF_T currpos = SMB_VFS_LSEEK(fsp, 0, SEEK_CUR);
unsigned char zero_space[4096];
SMB_OFF_T space_to_write;
if (currpos == -1)
return -1;
if (SMB_VFS_FSTAT(fsp, &st) == -1)
return -1;
space_to_write = len - st.st_size;
#ifdef S_ISFIFO
if (S_ISFIFO(st.st_mode))
return 0;
#endif
if (st.st_size == len)
return 0;
/* Shrink - just ftruncate. */
if (st.st_size > len)
return sys_ftruncate(fsp->fh->fd, len);
/* Write out the real space on disk. */
if (SMB_VFS_LSEEK(fsp, st.st_size, SEEK_SET) != st.st_size)
return -1;
space_to_write = len - st.st_size;
memset(zero_space, '\0', sizeof(zero_space));
while ( space_to_write > 0) {
SMB_OFF_T retlen;
SMB_OFF_T current_len_to_write = MIN(sizeof(zero_space),space_to_write);
retlen = SMB_VFS_WRITE(fsp,(char *)zero_space,current_len_to_write);
if (retlen <= 0)
return -1;
space_to_write -= retlen;
}
/* Seek to where we were */
if (SMB_VFS_LSEEK(fsp, currpos, SEEK_SET) != currpos)
return -1;
return 0;
}
static int vfswrap_ftruncate(vfs_handle_struct *handle, files_struct *fsp, SMB_OFF_T len)
{
int result = -1;
SMB_STRUCT_STAT st;
char c = 0;
SMB_OFF_T currpos;
START_PROFILE(syscall_ftruncate);
if (lp_strict_allocate(SNUM(fsp->conn))) {
result = strict_allocate_ftruncate(handle, fsp, len);
END_PROFILE(syscall_ftruncate);
return result;
}
/* we used to just check HAVE_FTRUNCATE_EXTEND and only use
sys_ftruncate if the system supports it. Then I discovered that
you can have some filesystems that support ftruncate
expansion and some that don't! On Linux fat can't do
ftruncate extend but ext2 can. */
result = sys_ftruncate(fsp->fh->fd, len);
if (result == 0)
goto done;
/* According to W. R. Stevens advanced UNIX prog. Pure 4.3 BSD cannot
extend a file with ftruncate. Provide alternate implementation
for this */
currpos = SMB_VFS_LSEEK(fsp, 0, SEEK_CUR);
if (currpos == -1) {
goto done;
}
/* Do an fstat to see if the file is longer than the requested
size in which case the ftruncate above should have
succeeded or shorter, in which case seek to len - 1 and
write 1 byte of zero */
if (SMB_VFS_FSTAT(fsp, &st) == -1) {
goto done;
}
#ifdef S_ISFIFO
if (S_ISFIFO(st.st_mode)) {
result = 0;
goto done;
}
#endif
if (st.st_size == len) {
result = 0;
goto done;
}
if (st.st_size > len) {
/* the sys_ftruncate should have worked */
goto done;
}
if (SMB_VFS_LSEEK(fsp, len-1, SEEK_SET) != len -1)
goto done;
if (SMB_VFS_WRITE(fsp, &c, 1)!=1)
goto done;
/* Seek to where we were */
if (SMB_VFS_LSEEK(fsp, currpos, SEEK_SET) != currpos)
goto done;
result = 0;
done:
END_PROFILE(syscall_ftruncate);
return result;
}
static bool vfswrap_lock(vfs_handle_struct *handle, files_struct *fsp, int op, SMB_OFF_T offset, SMB_OFF_T count, int type)
{
bool result;
START_PROFILE(syscall_fcntl_lock);
result = fcntl_lock(fsp->fh->fd, op, offset, count, type);
END_PROFILE(syscall_fcntl_lock);
return result;
}
static int vfswrap_kernel_flock(vfs_handle_struct *handle, files_struct *fsp,
uint32 share_mode)
{
START_PROFILE(syscall_kernel_flock);
kernel_flock(fsp->fh->fd, share_mode);
END_PROFILE(syscall_kernel_flock);
return 0;
}
static bool vfswrap_getlock(vfs_handle_struct *handle, files_struct *fsp, SMB_OFF_T *poffset, SMB_OFF_T *pcount, int *ptype, pid_t *ppid)
{
bool result;
START_PROFILE(syscall_fcntl_getlock);
result = fcntl_getlock(fsp->fh->fd, poffset, pcount, ptype, ppid);
END_PROFILE(syscall_fcntl_getlock);
return result;
}
static int vfswrap_linux_setlease(vfs_handle_struct *handle, files_struct *fsp,
int leasetype)
{
int result = -1;
START_PROFILE(syscall_linux_setlease);
#ifdef HAVE_KERNEL_OPLOCKS_LINUX
/* first set the signal handler */
if(linux_set_lease_sighandler(fsp->fh->fd) == -1) {
return -1;
}
result = linux_setlease(fsp->fh->fd, leasetype);
#else
errno = ENOSYS;
#endif
END_PROFILE(syscall_linux_setlease);
return result;
}
static int vfswrap_symlink(vfs_handle_struct *handle, const char *oldpath, const char *newpath)
{
int result;
START_PROFILE(syscall_symlink);
result = symlink(oldpath, newpath);
END_PROFILE(syscall_symlink);
return result;
}
static int vfswrap_readlink(vfs_handle_struct *handle, const char *path, char *buf, size_t bufsiz)
{
int result;
START_PROFILE(syscall_readlink);
result = readlink(path, buf, bufsiz);
END_PROFILE(syscall_readlink);
return result;
}
static int vfswrap_link(vfs_handle_struct *handle, const char *oldpath, const char *newpath)
{
int result;
START_PROFILE(syscall_link);
result = link(oldpath, newpath);
END_PROFILE(syscall_link);
return result;
}
static int vfswrap_mknod(vfs_handle_struct *handle, const char *pathname, mode_t mode, SMB_DEV_T dev)
{
int result;
START_PROFILE(syscall_mknod);
result = sys_mknod(pathname, mode, dev);
END_PROFILE(syscall_mknod);
return result;
}
static char *vfswrap_realpath(vfs_handle_struct *handle, const char *path, char *resolved_path)
{
char *result;
START_PROFILE(syscall_realpath);
result = realpath(path, resolved_path);
END_PROFILE(syscall_realpath);
return result;
}
static NTSTATUS vfswrap_notify_watch(vfs_handle_struct *vfs_handle,
struct sys_notify_context *ctx,
struct notify_entry *e,
void (*callback)(struct sys_notify_context *ctx,
void *private_data,
struct notify_event *ev),
void *private_data, void *handle)
{
/*
* So far inotify is the only supported default notify mechanism. If
* another platform like the the BSD's or a proprietary Unix comes
* along and wants another default, we can play the same trick we
* played with Posix ACLs.
*
* Until that is the case, hard-code inotify here.
*/
#ifdef HAVE_INOTIFY
if (lp_kernel_change_notify(ctx->conn->params)) {
return inotify_watch(ctx, e, callback, private_data, handle);
}
#endif
/*
* Do nothing, leave everything to notify_internal.c
*/
return NT_STATUS_OK;
}
static int vfswrap_chflags(vfs_handle_struct *handle, const char *path, int flags)
{
#ifdef HAVE_CHFLAGS
return chflags(path, flags);
#else
errno = ENOSYS;
return -1;
#endif
}
static struct file_id vfswrap_file_id_create(struct vfs_handle_struct *handle, SMB_DEV_T dev, SMB_INO_T inode)
{
return file_id_create_dev(dev, inode);
}
static NTSTATUS vfswrap_streaminfo(vfs_handle_struct *handle,
struct files_struct *fsp,
const char *fname,
TALLOC_CTX *mem_ctx,
unsigned int *pnum_streams,
struct stream_struct **pstreams)
{
SMB_STRUCT_STAT sbuf;
unsigned int num_streams = 0;
struct stream_struct *streams = NULL;
int ret;
if ((fsp != NULL) && (fsp->is_directory)) {
/*
* No default streams on directories
*/
goto done;
}
if ((fsp != NULL) && (fsp->fh->fd != -1)) {
ret = SMB_VFS_FSTAT(fsp, &sbuf);
}
else {
ret = SMB_VFS_STAT(handle->conn, fname, &sbuf);
}
if (ret == -1) {
return map_nt_error_from_unix(errno);
}
if (S_ISDIR(sbuf.st_mode)) {
goto done;
}
streams = talloc(mem_ctx, struct stream_struct);
if (streams == NULL) {
return NT_STATUS_NO_MEMORY;
}
streams->size = sbuf.st_size;
streams->alloc_size = get_allocation_size(handle->conn, fsp, &sbuf);
streams->name = talloc_strdup(streams, "::$DATA");
if (streams->name == NULL) {
TALLOC_FREE(streams);
return NT_STATUS_NO_MEMORY;
}
num_streams = 1;
done:
*pnum_streams = num_streams;
*pstreams = streams;
return NT_STATUS_OK;
}
static int vfswrap_get_real_filename(struct vfs_handle_struct *handle,
const char *path,
const char *name,
TALLOC_CTX *mem_ctx,
char **found_name)
{
return get_real_filename(handle->conn, path, name, mem_ctx,
found_name);
}
static NTSTATUS vfswrap_fget_nt_acl(vfs_handle_struct *handle,
files_struct *fsp,
uint32 security_info, SEC_DESC **ppdesc)
{
NTSTATUS result;
START_PROFILE(fget_nt_acl);
result = posix_fget_nt_acl(fsp, security_info, ppdesc);
END_PROFILE(fget_nt_acl);
return result;
}
static NTSTATUS vfswrap_get_nt_acl(vfs_handle_struct *handle,
const char *name,
uint32 security_info, SEC_DESC **ppdesc)
{
NTSTATUS result;
START_PROFILE(get_nt_acl);
result = posix_get_nt_acl(handle->conn, name, security_info, ppdesc);
END_PROFILE(get_nt_acl);
return result;
}
static NTSTATUS vfswrap_fset_nt_acl(vfs_handle_struct *handle, files_struct *fsp, uint32 security_info_sent, const SEC_DESC *psd)
{
NTSTATUS result;
START_PROFILE(fset_nt_acl);
result = set_nt_acl(fsp, security_info_sent, psd);
END_PROFILE(fset_nt_acl);
return result;
}
static int vfswrap_chmod_acl(vfs_handle_struct *handle, const char *name, mode_t mode)
{
#ifdef HAVE_NO_ACL
errno = ENOSYS;
return -1;
#else
int result;
START_PROFILE(chmod_acl);
result = chmod_acl(handle->conn, name, mode);
END_PROFILE(chmod_acl);
return result;
#endif
}
static int vfswrap_fchmod_acl(vfs_handle_struct *handle, files_struct *fsp, mode_t mode)
{
#ifdef HAVE_NO_ACL
errno = ENOSYS;
return -1;
#else
int result;
START_PROFILE(fchmod_acl);
result = fchmod_acl(fsp, mode);
END_PROFILE(fchmod_acl);
return result;
#endif
}
static int vfswrap_sys_acl_get_entry(vfs_handle_struct *handle, SMB_ACL_T theacl, int entry_id, SMB_ACL_ENTRY_T *entry_p)
{
return sys_acl_get_entry(theacl, entry_id, entry_p);
}
static int vfswrap_sys_acl_get_tag_type(vfs_handle_struct *handle, SMB_ACL_ENTRY_T entry_d, SMB_ACL_TAG_T *tag_type_p)
{
return sys_acl_get_tag_type(entry_d, tag_type_p);
}
static int vfswrap_sys_acl_get_permset(vfs_handle_struct *handle, SMB_ACL_ENTRY_T entry_d, SMB_ACL_PERMSET_T *permset_p)
{
return sys_acl_get_permset(entry_d, permset_p);
}
static void * vfswrap_sys_acl_get_qualifier(vfs_handle_struct *handle, SMB_ACL_ENTRY_T entry_d)
{
return sys_acl_get_qualifier(entry_d);
}
static SMB_ACL_T vfswrap_sys_acl_get_file(vfs_handle_struct *handle, const char *path_p, SMB_ACL_TYPE_T type)
{
return sys_acl_get_file(handle, path_p, type);
}
static SMB_ACL_T vfswrap_sys_acl_get_fd(vfs_handle_struct *handle, files_struct *fsp)
{
return sys_acl_get_fd(handle, fsp);
}
static int vfswrap_sys_acl_clear_perms(vfs_handle_struct *handle, SMB_ACL_PERMSET_T permset)
{
return sys_acl_clear_perms(permset);
}
static int vfswrap_sys_acl_add_perm(vfs_handle_struct *handle, SMB_ACL_PERMSET_T permset, SMB_ACL_PERM_T perm)
{
return sys_acl_add_perm(permset, perm);
}
static char * vfswrap_sys_acl_to_text(vfs_handle_struct *handle, SMB_ACL_T theacl, ssize_t *plen)
{
return sys_acl_to_text(theacl, plen);
}
static SMB_ACL_T vfswrap_sys_acl_init(vfs_handle_struct *handle, int count)
{
return sys_acl_init(count);
}
static int vfswrap_sys_acl_create_entry(vfs_handle_struct *handle, SMB_ACL_T *pacl, SMB_ACL_ENTRY_T *pentry)
{
return sys_acl_create_entry(pacl, pentry);
}
static int vfswrap_sys_acl_set_tag_type(vfs_handle_struct *handle, SMB_ACL_ENTRY_T entry, SMB_ACL_TAG_T tagtype)
{
return sys_acl_set_tag_type(entry, tagtype);
}
static int vfswrap_sys_acl_set_qualifier(vfs_handle_struct *handle, SMB_ACL_ENTRY_T entry, void *qual)
{
return sys_acl_set_qualifier(entry, qual);
}
static int vfswrap_sys_acl_set_permset(vfs_handle_struct *handle, SMB_ACL_ENTRY_T entry, SMB_ACL_PERMSET_T permset)
{
return sys_acl_set_permset(entry, permset);
}
static int vfswrap_sys_acl_valid(vfs_handle_struct *handle, SMB_ACL_T theacl )
{
return sys_acl_valid(theacl );
}
static int vfswrap_sys_acl_set_file(vfs_handle_struct *handle, const char *name, SMB_ACL_TYPE_T acltype, SMB_ACL_T theacl)
{
return sys_acl_set_file(handle, name, acltype, theacl);
}
static int vfswrap_sys_acl_set_fd(vfs_handle_struct *handle, files_struct *fsp, SMB_ACL_T theacl)
{
return sys_acl_set_fd(handle, fsp, theacl);
}
static int vfswrap_sys_acl_delete_def_file(vfs_handle_struct *handle, const char *path)
{
return sys_acl_delete_def_file(handle, path);
}
static int vfswrap_sys_acl_get_perm(vfs_handle_struct *handle, SMB_ACL_PERMSET_T permset, SMB_ACL_PERM_T perm)
{
return sys_acl_get_perm(permset, perm);
}
static int vfswrap_sys_acl_free_text(vfs_handle_struct *handle, char *text)
{
return sys_acl_free_text(text);
}
static int vfswrap_sys_acl_free_acl(vfs_handle_struct *handle, SMB_ACL_T posix_acl)
{
return sys_acl_free_acl(posix_acl);
}
static int vfswrap_sys_acl_free_qualifier(vfs_handle_struct *handle, void *qualifier, SMB_ACL_TAG_T tagtype)
{
return sys_acl_free_qualifier(qualifier, tagtype);
}
/****************************************************************
Extended attribute operations.
*****************************************************************/
static ssize_t vfswrap_getxattr(struct vfs_handle_struct *handle,const char *path, const char *name, void *value, size_t size)
{
return sys_getxattr(path, name, value, size);
}
static ssize_t vfswrap_lgetxattr(struct vfs_handle_struct *handle,const char *path, const char *name, void *value, size_t size)
{
return sys_lgetxattr(path, name, value, size);
}
static ssize_t vfswrap_fgetxattr(struct vfs_handle_struct *handle, struct files_struct *fsp, const char *name, void *value, size_t size)
{
return sys_fgetxattr(fsp->fh->fd, name, value, size);
}
static ssize_t vfswrap_listxattr(struct vfs_handle_struct *handle, const char *path, char *list, size_t size)
{
return sys_listxattr(path, list, size);
}
ssize_t vfswrap_llistxattr(struct vfs_handle_struct *handle, const char *path, char *list, size_t size)
{
return sys_llistxattr(path, list, size);
}
ssize_t vfswrap_flistxattr(struct vfs_handle_struct *handle, struct files_struct *fsp, char *list, size_t size)
{
return sys_flistxattr(fsp->fh->fd, list, size);
}
static int vfswrap_removexattr(struct vfs_handle_struct *handle, const char *path, const char *name)
{
return sys_removexattr(path, name);
}
static int vfswrap_lremovexattr(struct vfs_handle_struct *handle, const char *path, const char *name)
{
return sys_lremovexattr(path, name);
}
static int vfswrap_fremovexattr(struct vfs_handle_struct *handle, struct files_struct *fsp, const char *name)
{
return sys_fremovexattr(fsp->fh->fd, name);
}
static int vfswrap_setxattr(struct vfs_handle_struct *handle, const char *path, const char *name, const void *value, size_t size, int flags)
{
return sys_setxattr(path, name, value, size, flags);
}
static int vfswrap_lsetxattr(struct vfs_handle_struct *handle, const char *path, const char *name, const void *value, size_t size, int flags)
{
return sys_lsetxattr(path, name, value, size, flags);
}
static int vfswrap_fsetxattr(struct vfs_handle_struct *handle, struct files_struct *fsp, const char *name, const void *value, size_t size, int flags)
{
return sys_fsetxattr(fsp->fh->fd, name, value, size, flags);
}
static int vfswrap_aio_read(struct vfs_handle_struct *handle, struct files_struct *fsp, SMB_STRUCT_AIOCB *aiocb)
{
return sys_aio_read(aiocb);
}
static int vfswrap_aio_write(struct vfs_handle_struct *handle, struct files_struct *fsp, SMB_STRUCT_AIOCB *aiocb)
{
return sys_aio_write(aiocb);
}
static ssize_t vfswrap_aio_return(struct vfs_handle_struct *handle, struct files_struct *fsp, SMB_STRUCT_AIOCB *aiocb)
{
return sys_aio_return(aiocb);
}
static int vfswrap_aio_cancel(struct vfs_handle_struct *handle, struct files_struct *fsp, SMB_STRUCT_AIOCB *aiocb)
{
return sys_aio_cancel(fsp->fh->fd, aiocb);
}
static int vfswrap_aio_error(struct vfs_handle_struct *handle, struct files_struct *fsp, SMB_STRUCT_AIOCB *aiocb)
{
return sys_aio_error(aiocb);
}
static int vfswrap_aio_fsync(struct vfs_handle_struct *handle, struct files_struct *fsp, int op, SMB_STRUCT_AIOCB *aiocb)
{
return sys_aio_fsync(op, aiocb);
}
static int vfswrap_aio_suspend(struct vfs_handle_struct *handle, struct files_struct *fsp, const SMB_STRUCT_AIOCB * const aiocb[], int n, const struct timespec *timeout)
{
return sys_aio_suspend(aiocb, n, timeout);
}
static bool vfswrap_aio_force(struct vfs_handle_struct *handle, struct files_struct *fsp)
{
return false;
}
static bool vfswrap_is_offline(struct vfs_handle_struct *handle, const char *path, SMB_STRUCT_STAT *sbuf)
{
if (ISDOT(path) || ISDOTDOT(path)) {
return false;
}
if (!lp_dmapi_support(SNUM(handle->conn)) || !dmapi_have_session()) {
#if defined(ENOTSUP)
errno = ENOTSUP;
#endif
return false;
}
return (dmapi_file_flags(path) & FILE_ATTRIBUTE_OFFLINE) != 0;
}
static int vfswrap_set_offline(struct vfs_handle_struct *handle, const char *path)
{
/* We don't know how to set offline bit by default, needs to be overriden in the vfs modules */
#if defined(ENOTSUP)
errno = ENOTSUP;
#endif
return -1;
}
static vfs_op_tuple vfs_default_ops[] = {
/* Disk operations */
{SMB_VFS_OP(vfswrap_connect), SMB_VFS_OP_CONNECT,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_disconnect), SMB_VFS_OP_DISCONNECT,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_disk_free), SMB_VFS_OP_DISK_FREE,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_get_quota), SMB_VFS_OP_GET_QUOTA,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_set_quota), SMB_VFS_OP_SET_QUOTA,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_get_shadow_copy_data), SMB_VFS_OP_GET_SHADOW_COPY_DATA,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_statvfs), SMB_VFS_OP_STATVFS,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_fs_capabilities), SMB_VFS_OP_FS_CAPABILITIES,
SMB_VFS_LAYER_OPAQUE},
/* Directory operations */
{SMB_VFS_OP(vfswrap_opendir), SMB_VFS_OP_OPENDIR,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_readdir), SMB_VFS_OP_READDIR,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_seekdir), SMB_VFS_OP_SEEKDIR,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_telldir), SMB_VFS_OP_TELLDIR,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_rewinddir), SMB_VFS_OP_REWINDDIR,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_mkdir), SMB_VFS_OP_MKDIR,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_rmdir), SMB_VFS_OP_RMDIR,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_closedir), SMB_VFS_OP_CLOSEDIR,
SMB_VFS_LAYER_OPAQUE},
/* File operations */
{SMB_VFS_OP(vfswrap_open), SMB_VFS_OP_OPEN,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_create_file), SMB_VFS_OP_CREATE_FILE,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_close), SMB_VFS_OP_CLOSE,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_read), SMB_VFS_OP_READ,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_pread), SMB_VFS_OP_PREAD,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_write), SMB_VFS_OP_WRITE,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_pwrite), SMB_VFS_OP_PWRITE,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_lseek), SMB_VFS_OP_LSEEK,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_sendfile), SMB_VFS_OP_SENDFILE,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_recvfile), SMB_VFS_OP_RECVFILE,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_rename), SMB_VFS_OP_RENAME,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_fsync), SMB_VFS_OP_FSYNC,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_stat), SMB_VFS_OP_STAT,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_fstat), SMB_VFS_OP_FSTAT,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_lstat), SMB_VFS_OP_LSTAT,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_unlink), SMB_VFS_OP_UNLINK,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_chmod), SMB_VFS_OP_CHMOD,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_fchmod), SMB_VFS_OP_FCHMOD,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_chown), SMB_VFS_OP_CHOWN,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_fchown), SMB_VFS_OP_FCHOWN,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_lchown), SMB_VFS_OP_LCHOWN,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_chdir), SMB_VFS_OP_CHDIR,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_getwd), SMB_VFS_OP_GETWD,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_ntimes), SMB_VFS_OP_NTIMES,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_ftruncate), SMB_VFS_OP_FTRUNCATE,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_lock), SMB_VFS_OP_LOCK,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_kernel_flock), SMB_VFS_OP_KERNEL_FLOCK,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_linux_setlease), SMB_VFS_OP_LINUX_SETLEASE,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_getlock), SMB_VFS_OP_GETLOCK,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_symlink), SMB_VFS_OP_SYMLINK,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_readlink), SMB_VFS_OP_READLINK,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_link), SMB_VFS_OP_LINK,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_mknod), SMB_VFS_OP_MKNOD,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_realpath), SMB_VFS_OP_REALPATH,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_notify_watch), SMB_VFS_OP_NOTIFY_WATCH,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_chflags), SMB_VFS_OP_CHFLAGS,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_file_id_create), SMB_VFS_OP_FILE_ID_CREATE,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_streaminfo), SMB_VFS_OP_STREAMINFO,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_get_real_filename), SMB_VFS_OP_GET_REAL_FILENAME,
SMB_VFS_LAYER_OPAQUE},
/* NT ACL operations. */
{SMB_VFS_OP(vfswrap_fget_nt_acl), SMB_VFS_OP_FGET_NT_ACL,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_get_nt_acl), SMB_VFS_OP_GET_NT_ACL,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_fset_nt_acl), SMB_VFS_OP_FSET_NT_ACL,
SMB_VFS_LAYER_OPAQUE},
/* POSIX ACL operations. */
{SMB_VFS_OP(vfswrap_chmod_acl), SMB_VFS_OP_CHMOD_ACL,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_fchmod_acl), SMB_VFS_OP_FCHMOD_ACL,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_sys_acl_get_entry), SMB_VFS_OP_SYS_ACL_GET_ENTRY,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_sys_acl_get_tag_type), SMB_VFS_OP_SYS_ACL_GET_TAG_TYPE,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_sys_acl_get_permset), SMB_VFS_OP_SYS_ACL_GET_PERMSET,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_sys_acl_get_qualifier), SMB_VFS_OP_SYS_ACL_GET_QUALIFIER,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_sys_acl_get_file), SMB_VFS_OP_SYS_ACL_GET_FILE,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_sys_acl_get_fd), SMB_VFS_OP_SYS_ACL_GET_FD,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_sys_acl_clear_perms), SMB_VFS_OP_SYS_ACL_CLEAR_PERMS,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_sys_acl_add_perm), SMB_VFS_OP_SYS_ACL_ADD_PERM,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_sys_acl_to_text), SMB_VFS_OP_SYS_ACL_TO_TEXT,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_sys_acl_init), SMB_VFS_OP_SYS_ACL_INIT,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_sys_acl_create_entry), SMB_VFS_OP_SYS_ACL_CREATE_ENTRY,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_sys_acl_set_tag_type), SMB_VFS_OP_SYS_ACL_SET_TAG_TYPE,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_sys_acl_set_qualifier), SMB_VFS_OP_SYS_ACL_SET_QUALIFIER,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_sys_acl_set_permset), SMB_VFS_OP_SYS_ACL_SET_PERMSET,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_sys_acl_valid), SMB_VFS_OP_SYS_ACL_VALID,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_sys_acl_set_file), SMB_VFS_OP_SYS_ACL_SET_FILE,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_sys_acl_set_fd), SMB_VFS_OP_SYS_ACL_SET_FD,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_sys_acl_delete_def_file), SMB_VFS_OP_SYS_ACL_DELETE_DEF_FILE,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_sys_acl_get_perm), SMB_VFS_OP_SYS_ACL_GET_PERM,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_sys_acl_free_text), SMB_VFS_OP_SYS_ACL_FREE_TEXT,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_sys_acl_free_acl), SMB_VFS_OP_SYS_ACL_FREE_ACL,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_sys_acl_free_qualifier), SMB_VFS_OP_SYS_ACL_FREE_QUALIFIER,
SMB_VFS_LAYER_OPAQUE},
/* EA operations. */
{SMB_VFS_OP(vfswrap_getxattr), SMB_VFS_OP_GETXATTR,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_lgetxattr), SMB_VFS_OP_LGETXATTR,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_fgetxattr), SMB_VFS_OP_FGETXATTR,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_listxattr), SMB_VFS_OP_LISTXATTR,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_llistxattr), SMB_VFS_OP_LLISTXATTR,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_flistxattr), SMB_VFS_OP_FLISTXATTR,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_removexattr), SMB_VFS_OP_REMOVEXATTR,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_lremovexattr), SMB_VFS_OP_LREMOVEXATTR,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_fremovexattr), SMB_VFS_OP_FREMOVEXATTR,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_setxattr), SMB_VFS_OP_SETXATTR,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_lsetxattr), SMB_VFS_OP_LSETXATTR,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_fsetxattr), SMB_VFS_OP_FSETXATTR,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_aio_read), SMB_VFS_OP_AIO_READ,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_aio_write), SMB_VFS_OP_AIO_WRITE,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_aio_return), SMB_VFS_OP_AIO_RETURN,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_aio_cancel), SMB_VFS_OP_AIO_CANCEL,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_aio_error), SMB_VFS_OP_AIO_ERROR,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_aio_fsync), SMB_VFS_OP_AIO_FSYNC,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_aio_suspend),SMB_VFS_OP_AIO_SUSPEND,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_aio_force), SMB_VFS_OP_AIO_FORCE,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_is_offline),SMB_VFS_OP_IS_OFFLINE,
SMB_VFS_LAYER_OPAQUE},
{SMB_VFS_OP(vfswrap_set_offline),SMB_VFS_OP_SET_OFFLINE,
SMB_VFS_LAYER_OPAQUE},
/* Finish VFS operations definition */
{SMB_VFS_OP(NULL), SMB_VFS_OP_NOOP,
SMB_VFS_LAYER_NOOP}
};
NTSTATUS vfs_default_init(void);
NTSTATUS vfs_default_init(void)
{
unsigned int needed = SMB_VFS_OP_LAST + 1; /* convert from index to count */
if (ARRAY_SIZE(vfs_default_ops) != needed) {
DEBUG(0, ("%s: %u ops registered, but %u ops are required\n",
DEFAULT_VFS_MODULE_NAME, (unsigned int)ARRAY_SIZE(vfs_default_ops), needed));
smb_panic("operation(s) missing from default VFS module");
}
return smb_register_vfs(SMB_VFS_INTERFACE_VERSION,
DEFAULT_VFS_MODULE_NAME, vfs_default_ops);
}