along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
+/**
+ * @file vfs_glusterfs.c
+ * @author Anand Avati <avati@redhat.com>
+ * @date May 2013
+ * @brief Samba VFS module for glusterfs
+ *
+ * @todo
+ * - sendfile/recvfile support
+ *
+ * A Samba VFS module for GlusterFS, based on Gluster's libgfapi.
+ * This is a "bottom" vfs module (not something to be stacked on top of
+ * another module), and translates (most) calls to the closest actions
+ * available in libgfapi.
+ *
+ */
+
#include "includes.h"
#include "smbd/smbd.h"
#include <stdio.h>
#include "api/glfs.h"
#include "lib/util/dlinklist.h"
+#include "lib/util/tevent_unix.h"
+#ifdef HAVE_SYS_EVENTFD_H
+#include <sys/eventfd.h>
+#endif
+#include <pthread.h>
+#include "smbd/globals.h"
#define DEFAULT_VOLFILE_SERVER "localhost"
-/*
- TODO
- ----
- Short term:
- - AIO support
- - sendfile/recvfile support
-*/
-
-/* Helpers to provide 'integer' fds */
-
-/* This is global. gfapi's FD operations do not
- require filesystem context.
-*/
-
-static glfs_fd_t **glfd_fd;
-static int glfd_fd_size;
-static int glfd_fd_used;
-
-static int glfd_fd_store(glfs_fd_t *glfd)
-{
- int i;
- void *tmp;
-
- if (glfd_fd_size == glfd_fd_used) {
- if (glfd_fd_size >= INT_MAX - 1) {
- errno = ENOMEM;
- return -1;
- }
-
- tmp = talloc_realloc(glfd_fd, glfd_fd, glfs_fd_t *,
- glfd_fd_size + 1);
- if (tmp == NULL) {
- errno = ENOMEM;
- return -1;
- }
-
- glfd_fd = tmp;
- glfd_fd[glfd_fd_size] = 0;
- glfd_fd_size++;
- }
-
- for (i = 0; i < glfd_fd_size; i++) {
- if (glfd_fd[i] == NULL) {
- break;
- }
- }
- glfd_fd_used++;
- glfd_fd[i] = glfd;
- return i;
-}
-
-static glfs_fd_t *glfd_fd_get(int i)
-{
- if (i < 0 || i >= glfd_fd_size) {
- return NULL;
- }
- return glfd_fd[i];
-}
-
-static glfs_fd_t *glfd_fd_clear(int i)
-{
- glfs_fd_t *glfd = NULL;
-
- if (i < 0 || i >= glfd_fd_size) {
- return NULL;
- }
-
- glfd = glfd_fd[i];
-
- glfd_fd[i] = 0;
- glfd_fd_used--;
- return glfd;
-}
-
-/* Helper to convert stat to stat_ex */
+#ifdef HAVE_EVENTFD
+static pthread_mutex_t lock_req_list = PTHREAD_MUTEX_INITIALIZER;
+static int event_fd = -1;
+static struct tevent_fd *aio_read_event = NULL;
+static struct tevent_req **req_producer_list = NULL;
+static struct tevent_req **req_consumer_list = NULL;
+static uint64_t req_counter = 0;
+#endif
+/**
+ * Helper to convert struct stat to struct stat_ex.
+ */
static void smb_stat_ex_from_stat(struct stat_ex *dst, const struct stat *src)
{
ZERO_STRUCTP(dst);
dst->st_ex_rdev = src->st_rdev;
dst->st_ex_size = src->st_size;
dst->st_ex_atime.tv_sec = src->st_atime;
-#ifdef STAT_HAVE_NSEC
- dst->st_ex_atime.tv_nsec = src->st_atime_nsec;
-#endif
dst->st_ex_mtime.tv_sec = src->st_mtime;
-#ifdef STAT_HAVE_NSEC
- dst->st_ex_mtime.tv_nsec = src->st_mtime_nsec;
-#endif
dst->st_ex_ctime.tv_sec = src->st_ctime;
-#ifdef STAT_HAVE_NSEC
- dst->st_ex_ctime.tv_nsec = src->st_ctime_nsec;
-#endif
dst->st_ex_btime.tv_sec = src->st_mtime;
+ dst->st_ex_blksize = src->st_blksize;
+ dst->st_ex_blocks = src->st_blocks;
#ifdef STAT_HAVE_NSEC
+ dst->st_ex_atime.tv_nsec = src->st_atime_nsec;
+ dst->st_ex_mtime.tv_nsec = src->st_mtime_nsec;
+ dst->st_ex_ctime.tv_nsec = src->st_ctime_nsec;
dst->st_ex_btime.tv_nsec = src->st_mtime_nsec;
#endif
- dst->st_ex_blksize = src->st_blksize;
- dst->st_ex_blocks = src->st_blocks;
}
/* pre-opened glfs_t */
static struct glfs_preopened {
char *volume;
+ char *connectpath;
glfs_t *fs;
int ref;
struct glfs_preopened *next, *prev;
} *glfs_preopened;
-int glfs_set_preopened(const char *volume, glfs_t *fs)
+static int glfs_set_preopened(const char *volume, const char *connectpath, glfs_t *fs)
{
struct glfs_preopened *entry = NULL;
return -1;
}
+ entry->connectpath = talloc_strdup(entry, connectpath);
+ if (entry->connectpath == NULL) {
+ talloc_free(entry);
+ errno = ENOMEM;
+ return -1;
+ }
+
entry->fs = fs;
entry->ref = 1;
return 0;
}
-static glfs_t *glfs_find_preopened(const char *volume)
+static glfs_t *glfs_find_preopened(const char *volume, const char *connectpath)
{
struct glfs_preopened *entry = NULL;
for (entry = glfs_preopened; entry; entry = entry->next) {
- if (strcmp(entry->volume, volume) == 0) {
+ if (strcmp(entry->volume, volume) == 0 &&
+ strcmp(entry->connectpath, connectpath) == 0)
+ {
entry->ref++;
return entry->fs;
}
{
const char *volfile_server;
const char *volume;
- const char *logfile;
+ char *logfile;
int loglevel;
- glfs_t *fs;
- int ret;
+ glfs_t *fs = NULL;
+ TALLOC_CTX *tmp_ctx;
+ int ret = 0;
- logfile = lp_parm_const_string(SNUM(handle->conn), "glusterfs",
+ tmp_ctx = talloc_new(NULL);
+ if (tmp_ctx == NULL) {
+ ret = -1;
+ goto done;
+ }
+ logfile = lp_parm_talloc_string(tmp_ctx, SNUM(handle->conn), "glusterfs",
"logfile", NULL);
loglevel = lp_parm_int(SNUM(handle->conn), "glusterfs", "loglevel", -1);
volume = service;
}
- fs = glfs_find_preopened(volume);
+ fs = glfs_find_preopened(volume, handle->conn->connectpath);
if (fs) {
- goto found;
+ goto done;
}
fs = glfs_new(volume);
if (fs == NULL) {
- return -1;
+ ret = -1;
+ goto done;
}
ret = glfs_set_volfile_server(fs, "tcp", volfile_server, 0);
if (ret < 0) {
DEBUG(0, ("Failed to set volfile_server %s\n", volfile_server));
- glfs_fini(fs);
- return -1;
+ goto done;
}
ret = glfs_set_xlator_option(fs, "*-md-cache", "cache-posix-acl",
"true");
if (ret < 0) {
DEBUG(0, ("%s: Failed to set xlator options\n", volume));
+ goto done;
+ }
+
+
+ ret = glfs_set_xlator_option(fs, "*-snapview-client",
+ "snapdir-entry-path",
+ handle->conn->connectpath);
+ if (ret < 0) {
+ DEBUG(0, ("%s: Failed to set xlator option:"
+ " snapdir-entry-path\n", volume));
glfs_fini(fs);
return -1;
}
if (ret < 0) {
DEBUG(0, ("%s: Failed to set logfile %s loglevel %d\n",
volume, logfile, loglevel));
- glfs_fini(fs);
- return -1;
+ goto done;
}
ret = glfs_init(fs);
if (ret < 0) {
DEBUG(0, ("%s: Failed to initialize volume (%s)\n",
volume, strerror(errno)));
- glfs_fini(fs);
- return -1;
+ goto done;
}
- ret = glfs_set_preopened(volume, fs);
+ ret = glfs_set_preopened(volume, handle->conn->connectpath, fs);
if (ret < 0) {
DEBUG(0, ("%s: Failed to register volume (%s)\n",
volume, strerror(errno)));
- glfs_fini(fs);
+ goto done;
+ }
+done:
+ talloc_free(tmp_ctx);
+ if (ret < 0) {
+ if (fs)
+ glfs_fini(fs);
return -1;
+ } else {
+ DEBUG(0, ("%s: Initialized volume from server %s\n",
+ volume, volfile_server));
+ handle->data = fs;
+ return 0;
}
-found:
- DEBUG(0, ("%s: Initialized volume from server %s\n",
- volume, volfile_server));
- handle->data = fs;
- return 0;
}
static void vfs_gluster_disconnect(struct vfs_handle_struct *handle)
uint64_t *dsize_p)
{
struct statvfs statvfs = { 0, };
- uint64_t dfree = 0;
int ret;
ret = glfs_statvfs(handle->data, path, &statvfs);
return -1;
}
- dfree = statvfs.f_bsize * statvfs.f_bavail;
-
if (bsize_p != NULL) {
- *bsize_p = statvfs.f_bsize;
+ *bsize_p = (uint64_t)statvfs.f_bsize; /* Block size */
}
if (dfree_p != NULL) {
- *dfree_p = dfree;
+ *dfree_p = (uint64_t)statvfs.f_bavail; /* Available Block units */
}
if (dsize_p != NULL) {
- *dsize_p = statvfs.f_bsize * statvfs.f_blocks;
+ *dsize_p = (uint64_t)statvfs.f_blocks; /* Total Block units */
}
- return dfree;
+ return (uint64_t)statvfs.f_bavail;
}
static int vfs_gluster_get_quota(struct vfs_handle_struct *handle,
files_struct *fsp, const char *mask,
uint32 attributes)
{
- return (DIR *) glfd_fd_get(fsp->fh->fd);
+ return (DIR *) *(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp);
}
static int vfs_gluster_closedir(struct vfs_handle_struct *handle, DIR *dirp)
int flags, mode_t mode)
{
glfs_fd_t *glfd;
+ glfs_fd_t **p_tmp;
if (flags & O_DIRECTORY) {
glfd = glfs_opendir(handle->data, smb_fname->base_name);
}
if (glfd == NULL) {
- DEBUG(0, ("glfs_{open[dir],creat}(%s) failed: %s\n",
- smb_fname->base_name, strerror(errno)));
return -1;
}
-
- return glfd_fd_store(glfd);
+ p_tmp = (glfs_fd_t **)VFS_ADD_FSP_EXTENSION(handle, fsp,
+ glfs_fd_t *, NULL);
+ *p_tmp = glfd;
+ /* An arbitrary value for error reporting, so you know its us. */
+ return 13371337;
}
static int vfs_gluster_close(struct vfs_handle_struct *handle,
files_struct *fsp)
{
- return glfs_close(glfd_fd_clear(fsp->fh->fd));
+ glfs_fd_t *glfd;
+ glfd = *(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp);
+ VFS_REMOVE_FSP_EXTENSION(handle, fsp);
+ return glfs_close(glfd);
}
static ssize_t vfs_gluster_read(struct vfs_handle_struct *handle,
files_struct *fsp, void *data, size_t n)
{
- return glfs_read(glfd_fd_get(fsp->fh->fd), data, n, 0);
+ return glfs_read(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), data, n, 0);
}
static ssize_t vfs_gluster_pread(struct vfs_handle_struct *handle,
files_struct *fsp, void *data, size_t n,
off_t offset)
{
- return glfs_pread(glfd_fd_get(fsp->fh->fd), data, n, offset, 0);
+ return glfs_pread(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), data, n, offset, 0);
+}
+
+struct glusterfs_aio_state {
+ ssize_t ret;
+ int err;
+};
+
+/*
+ * This function is the callback that will be called on glusterfs
+ * threads once the async IO submitted is complete. To notify
+ * Samba of the completion we use eventfd mechanism.
+ */
+static void aio_glusterfs_done(glfs_fd_t *fd, ssize_t ret, void *data)
+{
+#if HAVE_EVENTFD
+ struct tevent_req *req = NULL;
+ struct glusterfs_aio_state *state = NULL;
+ int i, sts = 0;
+ uint64_t u = 1;
+
+ req = talloc_get_type_abort(data, struct tevent_req);
+ state = tevent_req_data(req, struct glusterfs_aio_state);
+
+ if (ret < 0) {
+ state->ret = -1;
+ state->err = errno;
+ } else {
+ state->ret = ret;
+ state->err = 0;
+ }
+
+ /*
+ * Store the reqs that needs to be completed by calling
+ * tevent_req_done(). tevent_req_done() cannot be called
+ * here, as it is not designed to be executed in the
+ * multithread environment, tevent_req_done() should be
+ * executed from the smbd main thread.
+ */
+ pthread_mutex_lock (&lock_req_list);
+ {
+ for (i = 0 ; i < aio_pending_size ; i++) {
+ if(!req_producer_list[i]) {
+ req_producer_list[i] = req;
+ req_counter = req_counter + 1;
+ break;
+ }
+ }
+ }
+ pthread_mutex_unlock (&lock_req_list);
+
+ /*
+ * For a bunch of fops notify only once
+ */
+ if (req_counter == 1) {
+ sts = write (event_fd, &u, sizeof(uint64_t));
+ if (sts < 0 && errno == EAGAIN)
+ DEBUG(0,("\nWRITE: reached max value"));
+ }
+ return;
+#endif
+}
+
+#ifdef HAVE_EVENTFD
+static void aio_tevent_fd_done(struct tevent_context *event_ctx,
+ struct tevent_fd *fde,
+ uint16 flags, void *data)
+{
+ struct tevent_req *req = NULL;
+ struct tevent_req **temp = NULL;
+ int i = 0, sts = 0;
+ uint64_t u = 0;
+
+ sts = read (event_fd, &u, sizeof(uint64_t));
+ if (sts < 0 && errno == EAGAIN)
+ DEBUG(0,("\nREAD: eventfd read failed (%s)",strerror(errno)));
+
+ pthread_mutex_lock (&lock_req_list);
+ {
+ temp = req_producer_list;
+ req_producer_list = req_consumer_list;
+ req_consumer_list = temp;
+ req_counter = 0;
+ }
+ pthread_mutex_unlock (&lock_req_list);
+
+ for (i = 0 ; i < aio_pending_size ; i++) {
+ req = req_consumer_list[i];
+ if (req) {
+ tevent_req_done(req);
+ req_consumer_list[i] = 0;
+ }
+ }
+ return;
+}
+#endif
+
+static bool init_gluster_aio(struct vfs_handle_struct *handle)
+{
+#ifdef HAVE_EVENTFD
+ if (event_fd != -1) {
+ /*
+ * Already initialized.
+ */
+ return true;
+ }
+
+ event_fd = eventfd(0, EFD_NONBLOCK | EFD_CLOEXEC);
+ if (event_fd == -1) {
+ goto fail;
+ }
+
+ aio_read_event = tevent_add_fd(handle->conn->sconn->ev_ctx,
+ NULL,
+ event_fd,
+ TEVENT_FD_READ,
+ aio_tevent_fd_done,
+ NULL);
+ if (aio_read_event == NULL) {
+ goto fail;
+ }
+
+ req_producer_list = talloc_zero_array(NULL, struct tevent_req *,
+ aio_pending_size);
+ req_consumer_list = talloc_zero_array(NULL, struct tevent_req *,
+ aio_pending_size);
+
+ return true;
+fail:
+ TALLOC_FREE(aio_read_event);
+ if (event_fd != -1) {
+ close(event_fd);
+ event_fd = -1;
+ }
+#endif
+ return false;
}
static struct tevent_req *vfs_gluster_pread_send(struct vfs_handle_struct
files_struct *fsp, void *data,
size_t n, off_t offset)
{
+ struct tevent_req *req = NULL;
+ struct glusterfs_aio_state *state = NULL;
+ int ret = 0;
+
+#ifndef HAVE_EVENTFD
errno = ENOTSUP;
return NULL;
-}
+#endif
-static ssize_t vfs_gluster_pread_recv(struct tevent_req *req, int *err)
-{
- errno = ENOTSUP;
- return -1;
+ req = tevent_req_create(mem_ctx, &state, struct glusterfs_aio_state);
+ if (req == NULL) {
+ return NULL;
+ }
+
+ if (!init_gluster_aio(handle)) {
+ tevent_req_error(req, EIO);
+ return tevent_req_post(req, ev);
+ }
+ ret = glfs_pread_async(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle,
+ fsp), data, n, offset, 0, aio_glusterfs_done,
+ req);
+ if (ret < 0) {
+ tevent_req_error(req, -ret);
+ return tevent_req_post(req, ev);
+ }
+
+ return req;
}
static ssize_t vfs_gluster_write(struct vfs_handle_struct *handle,
files_struct *fsp, const void *data, size_t n)
{
- return glfs_write(glfd_fd_get(fsp->fh->fd), data, n, 0);
+ return glfs_write(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), data, n, 0);
}
static ssize_t vfs_gluster_pwrite(struct vfs_handle_struct *handle,
files_struct *fsp, const void *data,
size_t n, off_t offset)
{
- return glfs_pwrite(glfd_fd_get(fsp->fh->fd), data, n, offset, 0);
+ return glfs_pwrite(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), data, n, offset, 0);
}
static struct tevent_req *vfs_gluster_pwrite_send(struct vfs_handle_struct
const void *data, size_t n,
off_t offset)
{
+ struct tevent_req *req = NULL;
+ struct glusterfs_aio_state *state = NULL;
+ int ret = 0;
+
+#ifndef HAVE_EVENTFD
errno = ENOTSUP;
return NULL;
+#endif
+
+ req = tevent_req_create(mem_ctx, &state, struct glusterfs_aio_state);
+ if (req == NULL) {
+ return NULL;
+ }
+ if (!init_gluster_aio(handle)) {
+ tevent_req_error(req, EIO);
+ return tevent_req_post(req, ev);
+ }
+ ret = glfs_pwrite_async(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle,
+ fsp), data, n, offset, 0, aio_glusterfs_done,
+ req);
+ if (ret < 0) {
+ tevent_req_error(req, -ret);
+ return tevent_req_post(req, ev);
+ }
+ return req;
}
-static ssize_t vfs_gluster_pwrite_recv(struct tevent_req *req, int *err)
+static ssize_t vfs_gluster_recv(struct tevent_req *req, int *err)
{
+ struct glusterfs_aio_state *state = NULL;
+
+#ifndef HAVE_EVENTFD
errno = ENOTSUP;
return -1;
+#endif
+ state = tevent_req_data(req, struct glusterfs_aio_state);
+ if (state == NULL) {
+ return -1;
+ }
+
+ if (tevent_req_is_unix_error(req, err)) {
+ return -1;
+ }
+ if (state->ret == -1) {
+ *err = state->err;
+ }
+ return state->ret;
}
static off_t vfs_gluster_lseek(struct vfs_handle_struct *handle,
files_struct *fsp, off_t offset, int whence)
{
- return glfs_lseek(glfd_fd_get(fsp->fh->fd), offset, whence);
+ return glfs_lseek(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), offset, whence);
}
static ssize_t vfs_gluster_sendfile(struct vfs_handle_struct *handle, int tofd,
static int vfs_gluster_fsync(struct vfs_handle_struct *handle,
files_struct *fsp)
{
- return glfs_fsync(glfd_fd_get(fsp->fh->fd));
+ return glfs_fsync(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp));
}
static struct tevent_req *vfs_gluster_fsync_send(struct vfs_handle_struct
struct tevent_context *ev,
files_struct *fsp)
{
+ struct tevent_req *req = NULL;
+ struct glusterfs_aio_state *state = NULL;
+ int ret = 0;
+
+#ifndef HAVE_EVENTFD
errno = ENOTSUP;
return NULL;
+#endif
+
+ req = tevent_req_create(mem_ctx, &state, struct glusterfs_aio_state);
+ if (req == NULL) {
+ return NULL;
+ }
+ if (!init_gluster_aio(handle)) {
+ tevent_req_error(req, EIO);
+ return tevent_req_post(req, ev);
+ }
+ ret = glfs_fsync_async(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle,
+ fsp), aio_glusterfs_done, req);
+ if (ret < 0) {
+ tevent_req_error(req, -ret);
+ return tevent_req_post(req, ev);
+ }
+ return req;
}
static int vfs_gluster_fsync_recv(struct tevent_req *req, int *err)
{
- errno = ENOTSUP;
- return -1;
+ /*
+ * Use implicit conversion ssize_t->int
+ */
+ return vfs_gluster_recv(req, err);
}
static int vfs_gluster_stat(struct vfs_handle_struct *handle,
struct stat st;
int ret;
- ret = glfs_fstat(glfd_fd_get(fsp->fh->fd), &st);
+ ret = glfs_fstat(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), &st);
if (ret == 0) {
smb_stat_ex_from_stat(sbuf, &st);
}
static int vfs_gluster_fchmod(struct vfs_handle_struct *handle,
files_struct *fsp, mode_t mode)
{
- return glfs_fchmod(glfd_fd_get(fsp->fh->fd), mode);
+ return glfs_fchmod(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), mode);
}
static int vfs_gluster_chown(struct vfs_handle_struct *handle,
static int vfs_gluster_fchown(struct vfs_handle_struct *handle,
files_struct *fsp, uid_t uid, gid_t gid)
{
- return glfs_fchown(glfd_fd_get(fsp->fh->fd), uid, gid);
+ return glfs_fchown(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), uid, gid);
}
static int vfs_gluster_lchown(struct vfs_handle_struct *handle,
char *cwd;
char *ret;
- cwd = calloc(1, PATH_MAX + 1);
+ cwd = SMB_CALLOC_ARRAY(char, PATH_MAX);
if (cwd == NULL) {
return NULL;
}
- ret = glfs_getcwd(handle->data, cwd, PATH_MAX);
+ ret = glfs_getcwd(handle->data, cwd, PATH_MAX - 1);
if (ret == 0) {
free(cwd);
}
{
struct timespec times[2];
- times[0].tv_sec = ft->atime.tv_sec;
- times[0].tv_nsec = ft->atime.tv_nsec;
- times[1].tv_sec = ft->mtime.tv_sec;
- times[1].tv_nsec = ft->mtime.tv_nsec;
+ if (null_timespec(ft->atime)) {
+ times[0].tv_sec = smb_fname->st.st_ex_atime.tv_sec;
+ times[0].tv_nsec = smb_fname->st.st_ex_atime.tv_nsec;
+ } else {
+ times[0].tv_sec = ft->atime.tv_sec;
+ times[0].tv_nsec = ft->atime.tv_nsec;
+ }
+
+ if (null_timespec(ft->mtime)) {
+ times[1].tv_sec = smb_fname->st.st_ex_mtime.tv_sec;
+ times[1].tv_nsec = smb_fname->st.st_ex_mtime.tv_nsec;
+ } else {
+ times[1].tv_sec = ft->mtime.tv_sec;
+ times[1].tv_nsec = ft->mtime.tv_nsec;
+ }
+
+ if ((timespec_compare(×[0],
+ &smb_fname->st.st_ex_atime) == 0) &&
+ (timespec_compare(×[1],
+ &smb_fname->st.st_ex_mtime) == 0)) {
+ return 0;
+ }
return glfs_utimens(handle->data, smb_fname->base_name, times);
}
static int vfs_gluster_ftruncate(struct vfs_handle_struct *handle,
files_struct *fsp, off_t offset)
{
- return glfs_ftruncate(glfd_fd_get(fsp->fh->fd), offset);
+ return glfs_ftruncate(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), offset);
}
static int vfs_gluster_fallocate(struct vfs_handle_struct *handle,
flock.l_len = count;
flock.l_pid = 0;
- ret = glfs_posix_lock(glfd_fd_get(fsp->fh->fd), op, &flock);
+ ret = glfs_posix_lock(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), op, &flock);
if (op == F_GETLK) {
/* lock query, true if someone else has locked */
flock.l_len = *pcount;
flock.l_pid = 0;
- ret = glfs_posix_lock(glfd_fd_get(fsp->fh->fd), F_GETLK, &flock);
+ ret = glfs_posix_lock(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), F_GETLK, &flock);
if (ret == -1) {
return false;
}
snprintf(key_buf, NAME_MAX + 64,
- "user.glusterfs.get_real_filename:%s", name);
+ "glusterfs.get_real_filename:%s", name);
ret = glfs_getxattr(handle->data, path, key_buf, val_buf, NAME_MAX + 1);
- if (ret == -1 && errno == ENODATA) {
- errno = EOPNOTSUPP;
+ if (ret == -1) {
+ if (errno == ENODATA) {
+ errno = EOPNOTSUPP;
+ }
return -1;
}
files_struct *fsp, const char *name,
void *value, size_t size)
{
- return glfs_fgetxattr(glfd_fd_get(fsp->fh->fd), name, value, size);
+ return glfs_fgetxattr(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), name, value, size);
}
static ssize_t vfs_gluster_listxattr(struct vfs_handle_struct *handle,
files_struct *fsp, char *list,
size_t size)
{
- return glfs_flistxattr(glfd_fd_get(fsp->fh->fd), list, size);
+ return glfs_flistxattr(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), list, size);
}
static int vfs_gluster_removexattr(struct vfs_handle_struct *handle,
static int vfs_gluster_fremovexattr(struct vfs_handle_struct *handle,
files_struct *fsp, const char *name)
{
- return glfs_fremovexattr(glfd_fd_get(fsp->fh->fd), name);
+ return glfs_fremovexattr(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), name);
}
static int vfs_gluster_setxattr(struct vfs_handle_struct *handle,
files_struct *fsp, const char *name,
const void *value, size_t size, int flags)
{
- return glfs_fsetxattr(glfd_fd_get(fsp->fh->fd), name, value, size,
+ return glfs_fsetxattr(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), name, value, size,
flags);
}
return -1;
}
-/* Posix ACL Operations */
+/*
+ Gluster ACL Format:
+
+ Size = 4 (header) + N * 8 (entry)
+
+ Offset Size Field (Little Endian)
+ -------------------------------------
+ 0-3 4-byte Version
+
+ 4-5 2-byte Entry-1 tag
+ 6-7 2-byte Entry-1 perm
+ 8-11 4-byte Entry-1 id
+
+ 12-13 2-byte Entry-2 tag
+ 14-15 2-byte Entry-2 perm
+ 16-19 4-byte Entry-2 id
+
+ ...
+ */
+
+/* header version */
#define GLUSTER_ACL_VERSION 2
+
+/* perm bits */
#define GLUSTER_ACL_READ 0x04
#define GLUSTER_ACL_WRITE 0x02
#define GLUSTER_ACL_EXECUTE 0x01
+/* tag values */
#define GLUSTER_ACL_UNDEFINED_TAG 0x00
#define GLUSTER_ACL_USER_OBJ 0x01
#define GLUSTER_ACL_USER 0x02
#define GLUSTER_ACL_UNDEFINED_ID (-1)
-struct gluster_ace {
- uint16_t tag;
- uint16_t perm;
- uint32_t id;
-};
+#define GLUSTER_ACL_HEADER_SIZE 4
+#define GLUSTER_ACL_ENTRY_SIZE 8
-struct gluster_acl_header {
- uint32_t version;
- struct gluster_ace entries[];
-};
+#define GLUSTER_ACL_SIZE(n) (GLUSTER_ACL_HEADER_SIZE + (n * GLUSTER_ACL_ENTRY_SIZE))
+
+static SMB_ACL_T mode_to_smb_acls(const struct stat *mode, TALLOC_CTX *mem_ctx)
+{
+ struct smb_acl_t *result;
+ int count;
+
+ count = 3;
+ result = sys_acl_init(mem_ctx);
+ if (!result) {
+ errno = ENOMEM;
+ return NULL;
+ }
+
+ result->acl = talloc_array(result, struct smb_acl_entry, count);
+ if (!result->acl) {
+ errno = ENOMEM;
+ talloc_free(result);
+ return NULL;
+ }
+
+ result->count = count;
+
+ result->acl[0].a_type = SMB_ACL_USER_OBJ;
+ result->acl[0].a_perm = (mode->st_mode & S_IRWXU) >> 6;;
+
+ result->acl[1].a_type = SMB_ACL_GROUP_OBJ;
+ result->acl[1].a_perm = (mode->st_mode & S_IRWXG) >> 3;;
+
+ result->acl[2].a_type = SMB_ACL_OTHER;
+ result->acl[2].a_perm = mode->st_mode & S_IRWXO;;
+
+ return result;
+}
static SMB_ACL_T gluster_to_smb_acl(const char *buf, size_t xattr_size,
TALLOC_CTX *mem_ctx)
{
int count;
size_t size;
- struct gluster_ace *ace;
struct smb_acl_entry *smb_ace;
- struct gluster_acl_header *hdr;
struct smb_acl_t *result;
int i;
+ int offset;
uint16_t tag;
uint16_t perm;
uint32_t id;
size = xattr_size;
- if (size < sizeof(*hdr)) {
- /* ACL should be at least as big as the header */
+ if (size < GLUSTER_ACL_HEADER_SIZE) {
+ /* ACL should be at least as big as the header (4 bytes) */
errno = EINVAL;
return NULL;
}
- size -= sizeof(*hdr);
+ size -= GLUSTER_ACL_HEADER_SIZE; /* size of header = 4 bytes */
- if (size % sizeof(*ace)) {
+ if (size % GLUSTER_ACL_ENTRY_SIZE) {
/* Size of entries must strictly be a multiple of
- size of an ACE
+ size of an ACE (8 bytes)
*/
errno = EINVAL;
return NULL;
}
- count = size / sizeof(*ace);
-
- hdr = (void *)buf;
+ count = size / GLUSTER_ACL_ENTRY_SIZE;
- if (ntohl(hdr->version) != GLUSTER_ACL_VERSION) {
+ /* Version is the first 4 bytes of the ACL */
+ if (IVAL(buf, 0) != GLUSTER_ACL_VERSION) {
DEBUG(0, ("Unknown gluster ACL version: %d\n",
- ntohl(hdr->version)));
+ IVAL(buf, 0)));
return NULL;
}
+ offset = GLUSTER_ACL_HEADER_SIZE;
result = sys_acl_init(mem_ctx);
if (!result) {
result->count = count;
smb_ace = result->acl;
- ace = hdr->entries;
for (i = 0; i < count; i++) {
- tag = ntohs(ace->tag);
+ /* TAG is the first 2 bytes of an entry */
+ tag = SVAL(buf, offset);
+ offset += 2;
+
+ /* PERM is the next 2 bytes of an entry */
+ perm = SVAL(buf, offset);
+ offset += 2;
+
+ /* ID is the last 4 bytes of an entry */
+ id = IVAL(buf, offset);
+ offset += 4;
switch(tag) {
case GLUSTER_ACL_USER:
return NULL;
}
- id = ntohl(ace->id);
switch(smb_ace->a_type) {
case SMB_ACL_USER:
break;
}
- perm = ntohs(ace->perm);
-
smb_ace->a_perm = 0;
smb_ace->a_perm |=
((perm & GLUSTER_ACL_READ) ? SMB_ACL_READ : 0);
smb_ace->a_perm |=
((perm & GLUSTER_ACL_EXECUTE) ? SMB_ACL_EXECUTE : 0);
- ace++;
smb_ace++;
}
}
+static int gluster_ace_cmp(const void *left, const void *right)
+{
+ int ret = 0;
+ uint16_t tag_left, tag_right;
+ uint32_t id_left, id_right;
+
+ /*
+ Sorting precedence:
+
+ - Smaller TAG values must be earlier.
+
+ - Within same TAG, smaller identifiers must be earlier, E.g:
+ UID 0 entry must be earlier than UID 200
+ GID 17 entry must be earlier than GID 19
+ */
+
+ /* TAG is the first element in the entry */
+ tag_left = SVAL(left, 0);
+ tag_right = SVAL(right, 0);
+
+ ret = (tag_left - tag_right);
+ if (!ret) {
+ /* ID is the third element in the entry, after two short
+ integers (tag and perm), i.e at offset 4.
+ */
+ id_left = IVAL(left, 4);
+ id_right = IVAL(right, 4);
+ ret = id_left - id_right;
+ }
+
+ return ret;
+}
+
+
static ssize_t smb_to_gluster_acl(SMB_ACL_T theacl, char *buf, size_t len)
{
ssize_t size;
- struct gluster_ace *ace;
struct smb_acl_entry *smb_ace;
- struct gluster_acl_header *hdr;
int i;
int count;
uint16_t tag;
uint16_t perm;
uint32_t id;
+ int offset;
count = theacl->count;
- size = sizeof(*hdr) + (count * sizeof(*ace));
+ size = GLUSTER_ACL_HEADER_SIZE + (count * GLUSTER_ACL_ENTRY_SIZE);
if (!buf) {
return size;
}
return -1;
}
- hdr = (void *)buf;
- ace = hdr->entries;
smb_ace = theacl->acl;
- hdr->version = htonl(GLUSTER_ACL_VERSION);
+ /* Version is the first 4 bytes of the ACL */
+ SIVAL(buf, 0, GLUSTER_ACL_VERSION);
+ offset = GLUSTER_ACL_HEADER_SIZE;
for (i = 0; i < count; i++) {
+ /* Calculate tag */
switch(smb_ace->a_type) {
case SMB_ACL_USER:
tag = GLUSTER_ACL_USER;
return -1;
}
- ace->tag = ntohs(tag);
+ /* Calculate id */
switch(smb_ace->a_type) {
case SMB_ACL_USER:
id = smb_ace->info.user.uid;
break;
}
- ace->id = ntohl(id);
+ /* Calculate perm */
+ perm = 0;
- ace->perm = 0;
- ace->perm |=
+ perm |=
((smb_ace->a_perm & SMB_ACL_READ) ? GLUSTER_ACL_READ : 0);
- ace->perm |=
+ perm |=
((smb_ace->a_perm & SMB_ACL_WRITE) ? GLUSTER_ACL_WRITE : 0);
- ace->perm |=
+ perm |=
((smb_ace->a_perm & SMB_ACL_EXECUTE) ? GLUSTER_ACL_EXECUTE : 0);
- ace++;
+
+ /* TAG is the first 2 bytes of an entry */
+ SSVAL(buf, offset, tag);
+ offset += 2;
+
+ /* PERM is the next 2 bytes of an entry */
+ SSVAL(buf, offset, perm);
+ offset += 2;
+
+ /* ID is the last 4 bytes of an entry */
+ SIVAL(buf, offset, id);
+ offset += 4;
+
smb_ace++;
}
+ /* Skip the header, sort @count number of 8-byte entries */
+ qsort(buf+GLUSTER_ACL_HEADER_SIZE, count, GLUSTER_ACL_ENTRY_SIZE,
+ gluster_ace_cmp);
+
return size;
}
TALLOC_CTX *mem_ctx)
{
struct smb_acl_t *result;
+ struct stat st;
char *buf;
- char *key;
- ssize_t ret;
+ const char *key;
+ ssize_t ret, size = GLUSTER_ACL_SIZE(20);
switch (type) {
case SMB_ACL_TYPE_ACCESS:
return NULL;
}
- ret = glfs_getxattr(handle->data, path_p, key, 0, 0);
- if (ret <= 0) {
+ buf = alloca(size);
+ if (!buf) {
return NULL;
}
- buf = alloca(ret);
- ret = glfs_getxattr(handle->data, path_p, key, buf, ret);
+ ret = glfs_getxattr(handle->data, path_p, key, buf, size);
+ if (ret == -1 && errno == ERANGE) {
+ ret = glfs_getxattr(handle->data, path_p, key, 0, 0);
+ if (ret > 0) {
+ buf = alloca(ret);
+ if (!buf) {
+ return NULL;
+ }
+ ret = glfs_getxattr(handle->data, path_p, key, buf, ret);
+ }
+ }
+
+ /* retrieving the ACL from the xattr has finally failed, do a
+ * mode-to-acl mapping */
+
+ if (ret == -1 && errno == ENODATA) {
+ ret = glfs_stat(handle->data, path_p, &st);
+ if (ret == 0) {
+ result = mode_to_smb_acls(&st, mem_ctx);
+ return result;
+ }
+ }
+
if (ret <= 0) {
return NULL;
}
TALLOC_CTX *mem_ctx)
{
struct smb_acl_t *result;
- int ret;
+ struct stat st;
+ ssize_t ret, size = GLUSTER_ACL_SIZE(20);
char *buf;
+ glfs_fd_t *glfd;
- ret = glfs_fgetxattr(glfd_fd_get(fsp->fh->fd),
- "system.posix_acl_access", 0, 0);
- if (ret <= 0) {
+ glfd = *(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp);
+
+ buf = alloca(size);
+ if (!buf) {
return NULL;
}
- buf = alloca(ret);
- ret = glfs_fgetxattr(glfd_fd_get(fsp->fh->fd),
- "system.posix_acl_access", buf, ret);
+ ret = glfs_fgetxattr(glfd, "system.posix_acl_access", buf, size);
+ if (ret == -1 && errno == ERANGE) {
+ ret = glfs_fgetxattr(glfd, "system.posix_acl_access", 0, 0);
+ if (ret > 0) {
+ buf = alloca(ret);
+ if (!buf) {
+ return NULL;
+ }
+ ret = glfs_fgetxattr(glfd, "system.posix_acl_access",
+ buf, ret);
+ }
+ }
+
+ /* retrieving the ACL from the xattr has finally failed, do a
+ * mode-to-acl mapping */
+
+ if (ret == -1 && errno == ENODATA) {
+ ret = glfs_fstat(glfd, &st);
+ if (ret == 0) {
+ result = mode_to_smb_acls(&st, mem_ctx);
+ return result;
+ }
+ }
+
if (ret <= 0) {
return NULL;
}
SMB_ACL_T theacl)
{
int ret;
- char *key;
+ const char *key;
char *buf;
ssize_t size;
SMB_ACL_T theacl)
{
int ret;
- char *key;
char *buf;
ssize_t size;
return -1;
}
- ret = glfs_fsetxattr(glfd_fd_get(fsp->fh->fd),
+ ret = glfs_fsetxattr(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp),
"system.posix_acl_access", buf, size, 0);
return ret;
}
.read_fn = vfs_gluster_read,
.pread_fn = vfs_gluster_pread,
.pread_send_fn = vfs_gluster_pread_send,
- .pread_recv_fn = vfs_gluster_pread_recv,
+ .pread_recv_fn = vfs_gluster_recv,
.write_fn = vfs_gluster_write,
.pwrite_fn = vfs_gluster_pwrite,
.pwrite_send_fn = vfs_gluster_pwrite_send,
- .pwrite_recv_fn = vfs_gluster_pwrite_recv,
+ .pwrite_recv_fn = vfs_gluster_recv,
.lseek_fn = vfs_gluster_lseek,
.sendfile_fn = vfs_gluster_sendfile,
.recvfile_fn = vfs_gluster_recvfile,
git.samba.org / obnox / samba / samba-obnox.git / blobdiff