auth/credentials: don't ignore "client use kerberos" and --use-kerberos for machine...
[Samba.git] / source3 / modules / vfs_ceph.c
blobbb1650446c9669879033ab3cba6dd5b2d1ca7e6a
1 /*
2 Unix SMB/CIFS implementation.
3 Wrap disk only vfs functions to sidestep dodgy compilers.
4 Copyright (C) Tim Potter 1998
5 Copyright (C) Jeremy Allison 2007
6 Copyright (C) Brian Chrisman 2011 <bchrisman@gmail.com>
7 Copyright (C) Richard Sharpe 2011 <realrichardsharpe@gmail.com>
9 This program is free software; you can redistribute it and/or modify
10 it under the terms of the GNU General Public License as published by
11 the Free Software Foundation; either version 3 of the License, or
12 (at your option) any later version.
14 This program is distributed in the hope that it will be useful,
15 but WITHOUT ANY WARRANTY; without even the implied warranty of
16 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17 GNU General Public License for more details.
19 You should have received a copy of the GNU General Public License
20 along with this program. If not, see <http://www.gnu.org/licenses/>.
24 * This VFS only works with the libcephfs.so user-space client. It is not needed
25 * if you are using the kernel client or the FUSE client.
27 * Add the following smb.conf parameter to each share that will be hosted on
28 * Ceph:
30 * vfs objects = [any others you need go here] ceph
33 #include "includes.h"
34 #include "smbd/smbd.h"
35 #include "system/filesys.h"
36 #include <dirent.h>
37 #include <sys/statvfs.h>
38 #include "cephfs/libcephfs.h"
39 #include "smbprofile.h"
40 #include "modules/posixacl_xattr.h"
41 #include "lib/util/tevent_unix.h"
43 #undef DBGC_CLASS
44 #define DBGC_CLASS DBGC_VFS
46 #ifndef LIBCEPHFS_VERSION
47 #define LIBCEPHFS_VERSION(maj, min, extra) ((maj << 16) + (min << 8) + extra)
48 #define LIBCEPHFS_VERSION_CODE LIBCEPHFS_VERSION(0, 0, 0)
49 #endif
52 * Use %llu whenever we have a 64bit unsigned int, and cast to (long long unsigned)
54 #define llu(_var) ((long long unsigned)_var)
57 * Note, libcephfs's return code model is to return -errno! So we have to
58 * convert to what Samba expects, which is to set errno to -return and return -1
60 #define WRAP_RETURN(_res) \
61 errno = 0; \
62 if (_res < 0) { \
63 errno = -_res; \
64 return -1; \
65 } \
66 return _res \
69 * Track unique connections, as virtual mounts, to cephfs file systems.
70 * Individual mounts will be set on the handle->data attribute, but
71 * the mounts themselves will be shared so as not to spawn extra mounts
72 * to the same cephfs.
74 * Individual mounts are IDed by a 'cookie' value that is a string built
75 * from identifying parameters found in smb.conf.
78 static struct cephmount_cached {
79 char *cookie;
80 uint32_t count;
81 struct ceph_mount_info *mount;
82 struct cephmount_cached *next, *prev;
83 } *cephmount_cached;
85 static int cephmount_cache_add(const char *cookie,
86 struct ceph_mount_info *mount)
88 struct cephmount_cached *entry = NULL;
90 entry = talloc_zero(NULL, struct cephmount_cached);
91 if (entry == NULL) {
92 errno = ENOMEM;
93 return -1;
96 entry->cookie = talloc_strdup(entry, cookie);
97 if (entry->cookie == NULL) {
98 talloc_free(entry);
99 errno = ENOMEM;
100 return -1;
103 entry->mount = mount;
104 entry->count = 1;
106 DBG_DEBUG("adding mount cache entry for %s\n", entry->cookie);
107 DLIST_ADD(cephmount_cached, entry);
108 return 0;
111 static struct ceph_mount_info *cephmount_cache_update(const char *cookie)
113 struct cephmount_cached *entry = NULL;
115 for (entry = cephmount_cached; entry; entry = entry->next) {
116 if (strcmp(entry->cookie, cookie) == 0) {
117 entry->count++;
118 DBG_DEBUG("updated mount cache: count is [%"
119 PRIu32 "]\n", entry->count);
120 return entry->mount;
124 errno = ENOENT;
125 return NULL;
128 static int cephmount_cache_remove(struct ceph_mount_info *mount)
130 struct cephmount_cached *entry = NULL;
132 for (entry = cephmount_cached; entry; entry = entry->next) {
133 if (entry->mount == mount) {
134 if (--entry->count) {
135 DBG_DEBUG("updated mount cache: count is [%"
136 PRIu32 "]\n", entry->count);
137 return entry->count;
140 DBG_DEBUG("removing mount cache entry for %s\n",
141 entry->cookie);
142 DLIST_REMOVE(cephmount_cached, entry);
143 talloc_free(entry);
144 return 0;
147 errno = ENOENT;
148 return -1;
151 static char *cephmount_get_cookie(TALLOC_CTX * mem_ctx, const int snum)
153 const char *conf_file =
154 lp_parm_const_string(snum, "ceph", "config_file", ".");
155 const char *user_id = lp_parm_const_string(snum, "ceph", "user_id", "");
156 const char *fsname =
157 lp_parm_const_string(snum, "ceph", "filesystem", "");
158 return talloc_asprintf(mem_ctx, "(%s/%s/%s)", conf_file, user_id,
159 fsname);
162 static int cephmount_select_fs(struct ceph_mount_info *mnt, const char *fsname)
165 * ceph_select_filesystem was added in ceph 'nautilus' (v14).
166 * Earlier versions of libcephfs will lack that API function.
167 * At the time of this writing (Feb 2023) all versions of ceph
168 * supported by ceph upstream have this function.
170 #if defined(HAVE_CEPH_SELECT_FILESYSTEM)
171 DBG_DEBUG("[CEPH] calling: ceph_select_filesystem with %s\n", fsname);
172 return ceph_select_filesystem(mnt, fsname);
173 #else
174 DBG_ERR("[CEPH] ceph_select_filesystem not available\n");
175 return -ENOTSUP;
176 #endif
179 static struct ceph_mount_info *cephmount_mount_fs(const int snum)
181 int ret;
182 char buf[256];
183 struct ceph_mount_info *mnt = NULL;
184 /* if config_file and/or user_id are NULL, ceph will use defaults */
185 const char *conf_file =
186 lp_parm_const_string(snum, "ceph", "config_file", NULL);
187 const char *user_id =
188 lp_parm_const_string(snum, "ceph", "user_id", NULL);
189 const char *fsname =
190 lp_parm_const_string(snum, "ceph", "filesystem", NULL);
192 DBG_DEBUG("[CEPH] calling: ceph_create\n");
193 ret = ceph_create(&mnt, user_id);
194 if (ret) {
195 errno = -ret;
196 return NULL;
199 DBG_DEBUG("[CEPH] calling: ceph_conf_read_file with %s\n",
200 (conf_file == NULL ? "default path" : conf_file));
201 ret = ceph_conf_read_file(mnt, conf_file);
202 if (ret) {
203 goto err_cm_release;
206 DBG_DEBUG("[CEPH] calling: ceph_conf_get\n");
207 ret = ceph_conf_get(mnt, "log file", buf, sizeof(buf));
208 if (ret < 0) {
209 goto err_cm_release;
212 /* libcephfs disables POSIX ACL support by default, enable it... */
213 ret = ceph_conf_set(mnt, "client_acl_type", "posix_acl");
214 if (ret < 0) {
215 goto err_cm_release;
217 /* tell libcephfs to perform local permission checks */
218 ret = ceph_conf_set(mnt, "fuse_default_permissions", "false");
219 if (ret < 0) {
220 goto err_cm_release;
223 * select a cephfs file system to use:
224 * In ceph, multiple file system support has been stable since 'pacific'.
225 * Permit different shares to access different file systems.
227 if (fsname != NULL) {
228 ret = cephmount_select_fs(mnt, fsname);
229 if (ret < 0) {
230 goto err_cm_release;
234 DBG_DEBUG("[CEPH] calling: ceph_mount\n");
235 ret = ceph_mount(mnt, NULL);
236 if (ret >= 0) {
237 goto cm_done;
240 err_cm_release:
241 ceph_release(mnt);
242 mnt = NULL;
243 DBG_DEBUG("[CEPH] Error mounting fs: %s\n", strerror(-ret));
244 cm_done:
246 * Handle the error correctly. Ceph returns -errno.
248 if (ret) {
249 errno = -ret;
251 return mnt;
254 /* Check for NULL pointer parameters in cephwrap_* functions */
256 /* We don't want to have NULL function pointers lying around. Someone
257 is sure to try and execute them. These stubs are used to prevent
258 this possibility. */
260 static int cephwrap_connect(struct vfs_handle_struct *handle,
261 const char *service, const char *user)
263 int ret = 0;
264 struct ceph_mount_info *cmount = NULL;
265 int snum = SNUM(handle->conn);
266 char *cookie = cephmount_get_cookie(handle, snum);
267 if (cookie == NULL) {
268 return -1;
271 cmount = cephmount_cache_update(cookie);
272 if (cmount != NULL) {
273 goto connect_ok;
276 cmount = cephmount_mount_fs(snum);
277 if (cmount == NULL) {
278 ret = -1;
279 goto connect_fail;
281 ret = cephmount_cache_add(cookie, cmount);
282 if (ret) {
283 goto connect_fail;
286 connect_ok:
287 handle->data = cmount;
288 DBG_WARNING("Connection established with the server: %s\n", cookie);
290 * Unless we have an async implementation of getxattrat turn this off.
292 lp_do_parameter(SNUM(handle->conn), "smbd async dosmode", "false");
293 connect_fail:
294 talloc_free(cookie);
295 return ret;
298 static void cephwrap_disconnect(struct vfs_handle_struct *handle)
300 int ret = cephmount_cache_remove(handle->data);
301 if (ret < 0) {
302 DBG_ERR("failed to remove ceph mount from cache: %s\n",
303 strerror(errno));
304 return;
306 if (ret > 0) {
307 DBG_DEBUG("mount cache entry still in use\n");
308 return;
311 ret = ceph_unmount(handle->data);
312 if (ret < 0) {
313 DBG_ERR("[CEPH] failed to unmount: %s\n", strerror(-ret));
316 ret = ceph_release(handle->data);
317 if (ret < 0) {
318 DBG_ERR("[CEPH] failed to release: %s\n", strerror(-ret));
320 handle->data = NULL;
323 /* Disk operations */
325 static uint64_t cephwrap_disk_free(struct vfs_handle_struct *handle,
326 const struct smb_filename *smb_fname,
327 uint64_t *bsize,
328 uint64_t *dfree,
329 uint64_t *dsize)
331 struct statvfs statvfs_buf = { 0 };
332 int ret;
334 if (!(ret = ceph_statfs(handle->data, smb_fname->base_name,
335 &statvfs_buf))) {
337 * Provide all the correct values.
339 *bsize = statvfs_buf.f_bsize;
340 *dfree = statvfs_buf.f_bavail;
341 *dsize = statvfs_buf.f_blocks;
342 DBG_DEBUG("[CEPH] bsize: %llu, dfree: %llu, dsize: %llu\n",
343 llu(*bsize), llu(*dfree), llu(*dsize));
344 return *dfree;
345 } else {
346 DBG_DEBUG("[CEPH] ceph_statfs returned %d\n", ret);
347 WRAP_RETURN(ret);
351 static int cephwrap_get_quota(struct vfs_handle_struct *handle,
352 const struct smb_filename *smb_fname,
353 enum SMB_QUOTA_TYPE qtype,
354 unid_t id,
355 SMB_DISK_QUOTA *qt)
357 /* libcephfs: Ceph does not implement this */
358 #if 0
359 /* was ifdef HAVE_SYS_QUOTAS */
360 int ret;
362 ret = ceph_get_quota(handle->conn->connectpath, qtype, id, qt);
364 if (ret) {
365 errno = -ret;
366 ret = -1;
369 return ret;
370 #else
371 errno = ENOSYS;
372 return -1;
373 #endif
376 static int cephwrap_set_quota(struct vfs_handle_struct *handle, enum SMB_QUOTA_TYPE qtype, unid_t id, SMB_DISK_QUOTA *qt)
378 /* libcephfs: Ceph does not implement this */
379 #if 0
380 /* was ifdef HAVE_SYS_QUOTAS */
381 int ret;
383 ret = ceph_set_quota(handle->conn->connectpath, qtype, id, qt);
384 if (ret) {
385 errno = -ret;
386 ret = -1;
389 return ret;
390 #else
391 WRAP_RETURN(-ENOSYS);
392 #endif
395 static int cephwrap_statvfs(struct vfs_handle_struct *handle,
396 const struct smb_filename *smb_fname,
397 struct vfs_statvfs_struct *statbuf)
399 struct statvfs statvfs_buf = { 0 };
400 int ret;
402 ret = ceph_statfs(handle->data, smb_fname->base_name, &statvfs_buf);
403 if (ret < 0) {
404 WRAP_RETURN(ret);
407 statbuf->OptimalTransferSize = statvfs_buf.f_frsize;
408 statbuf->BlockSize = statvfs_buf.f_bsize;
409 statbuf->TotalBlocks = statvfs_buf.f_blocks;
410 statbuf->BlocksAvail = statvfs_buf.f_bfree;
411 statbuf->UserBlocksAvail = statvfs_buf.f_bavail;
412 statbuf->TotalFileNodes = statvfs_buf.f_files;
413 statbuf->FreeFileNodes = statvfs_buf.f_ffree;
414 statbuf->FsIdentifier = statvfs_buf.f_fsid;
415 DBG_DEBUG("[CEPH] f_bsize: %ld, f_blocks: %ld, f_bfree: %ld, f_bavail: %ld\n",
416 (long int)statvfs_buf.f_bsize, (long int)statvfs_buf.f_blocks,
417 (long int)statvfs_buf.f_bfree, (long int)statvfs_buf.f_bavail);
419 return ret;
422 static uint32_t cephwrap_fs_capabilities(struct vfs_handle_struct *handle,
423 enum timestamp_set_resolution *p_ts_res)
425 uint32_t caps = FILE_CASE_SENSITIVE_SEARCH | FILE_CASE_PRESERVED_NAMES;
427 *p_ts_res = TIMESTAMP_SET_NT_OR_BETTER;
429 return caps;
432 /* Directory operations */
434 static DIR *cephwrap_fdopendir(struct vfs_handle_struct *handle,
435 struct files_struct *fsp,
436 const char *mask,
437 uint32_t attributes)
439 int ret = 0;
440 struct ceph_dir_result *result = NULL;
442 #ifdef HAVE_CEPH_FDOPENDIR
443 int dirfd = fsp_get_io_fd(fsp);
444 DBG_DEBUG("[CEPH] fdopendir(%p, %d)\n", handle, dirfd);
445 ret = ceph_fdopendir(handle->data, dirfd, &result);
446 #else
447 DBG_DEBUG("[CEPH] fdopendir(%p, %p)\n", handle, fsp);
448 ret = ceph_opendir(handle->data, fsp->fsp_name->base_name, &result);
449 #endif
450 if (ret < 0) {
451 result = NULL;
452 errno = -ret; /* We return result which is NULL in this case */
455 DBG_DEBUG("[CEPH] fdopendir(...) = %d\n", ret);
456 return (DIR *) result;
459 static struct dirent *cephwrap_readdir(struct vfs_handle_struct *handle,
460 struct files_struct *dirfsp,
461 DIR *dirp)
463 struct dirent *result = NULL;
465 DBG_DEBUG("[CEPH] readdir(%p, %p)\n", handle, dirp);
466 result = ceph_readdir(handle->data, (struct ceph_dir_result *) dirp);
467 DBG_DEBUG("[CEPH] readdir(...) = %p\n", result);
469 return result;
472 static void cephwrap_rewinddir(struct vfs_handle_struct *handle, DIR *dirp)
474 DBG_DEBUG("[CEPH] rewinddir(%p, %p)\n", handle, dirp);
475 ceph_rewinddir(handle->data, (struct ceph_dir_result *) dirp);
478 static int cephwrap_mkdirat(struct vfs_handle_struct *handle,
479 files_struct *dirfsp,
480 const struct smb_filename *smb_fname,
481 mode_t mode)
483 int result = -1;
484 #ifdef HAVE_CEPH_MKDIRAT
485 int dirfd = fsp_get_pathref_fd(dirfsp);
487 DBG_DEBUG("[CEPH] mkdirat(%p, %d, %s)\n",
488 handle,
489 dirfd,
490 smb_fname->base_name);
492 result = ceph_mkdirat(handle->data, dirfd, smb_fname->base_name, mode);
494 DBG_DEBUG("[CEPH] mkdirat(...) = %d\n", result);
496 WRAP_RETURN(result);
497 #else
498 struct smb_filename *full_fname = NULL;
500 full_fname = full_path_from_dirfsp_atname(talloc_tos(),
501 dirfsp,
502 smb_fname);
503 if (full_fname == NULL) {
504 return -1;
507 DBG_DEBUG("[CEPH] mkdir(%p, %s)\n",
508 handle, smb_fname_str_dbg(full_fname));
510 result = ceph_mkdir(handle->data, full_fname->base_name, mode);
512 TALLOC_FREE(full_fname);
514 WRAP_RETURN(result);
515 #endif
518 static int cephwrap_closedir(struct vfs_handle_struct *handle, DIR *dirp)
520 int result;
522 DBG_DEBUG("[CEPH] closedir(%p, %p)\n", handle, dirp);
523 result = ceph_closedir(handle->data, (struct ceph_dir_result *) dirp);
524 DBG_DEBUG("[CEPH] closedir(...) = %d\n", result);
525 WRAP_RETURN(result);
528 /* File operations */
530 static int cephwrap_openat(struct vfs_handle_struct *handle,
531 const struct files_struct *dirfsp,
532 const struct smb_filename *smb_fname,
533 files_struct *fsp,
534 const struct vfs_open_how *how)
536 int flags = how->flags;
537 mode_t mode = how->mode;
538 struct smb_filename *name = NULL;
539 bool have_opath = false;
540 bool became_root = false;
541 int result = -ENOENT;
542 #ifdef HAVE_CEPH_OPENAT
543 int dirfd = -1;
544 #endif
546 if (how->resolve != 0) {
547 errno = ENOSYS;
548 return -1;
551 if (smb_fname->stream_name) {
552 goto out;
555 #ifdef O_PATH
556 have_opath = true;
557 if (fsp->fsp_flags.is_pathref) {
558 flags |= O_PATH;
560 #endif
562 #ifdef HAVE_CEPH_OPENAT
563 dirfd = fsp_get_pathref_fd(dirfsp);
565 DBG_DEBUG("[CEPH] openat(%p, %d, %p, %d, %d)\n",
566 handle, dirfd, fsp, flags, mode);
568 if (fsp->fsp_flags.is_pathref && !have_opath) {
569 become_root();
570 became_root = true;
573 result = ceph_openat(handle->data,
574 dirfd,
575 smb_fname->base_name,
576 flags,
577 mode);
579 #else
580 if (fsp_get_pathref_fd(dirfsp) != AT_FDCWD) {
581 name = full_path_from_dirfsp_atname(talloc_tos(),
582 dirfsp,
583 smb_fname);
584 if (name == NULL) {
585 return -1;
587 smb_fname = name;
590 DBG_DEBUG("[CEPH] openat(%p, %s, %p, %d, %d)\n", handle,
591 smb_fname_str_dbg(smb_fname), fsp, flags, mode);
593 if (fsp->fsp_flags.is_pathref && !have_opath) {
594 become_root();
595 became_root = true;
598 result = ceph_open(handle->data, smb_fname->base_name, flags, mode);
599 #endif
600 if (became_root) {
601 unbecome_root();
603 out:
604 TALLOC_FREE(name);
605 fsp->fsp_flags.have_proc_fds = false;
606 DBG_DEBUG("[CEPH] open(...) = %d\n", result);
607 WRAP_RETURN(result);
610 static int cephwrap_close(struct vfs_handle_struct *handle, files_struct *fsp)
612 int result;
614 DBG_DEBUG("[CEPH] close(%p, %p)\n", handle, fsp);
615 result = ceph_close(handle->data, fsp_get_pathref_fd(fsp));
616 DBG_DEBUG("[CEPH] close(...) = %d\n", result);
618 WRAP_RETURN(result);
621 static ssize_t cephwrap_pread(struct vfs_handle_struct *handle, files_struct *fsp, void *data,
622 size_t n, off_t offset)
624 ssize_t result;
626 DBG_DEBUG("[CEPH] pread(%p, %p, %p, %llu, %llu)\n", handle, fsp, data, llu(n), llu(offset));
628 result = ceph_read(handle->data, fsp_get_io_fd(fsp), data, n, offset);
629 DBG_DEBUG("[CEPH] pread(...) = %llu\n", llu(result));
630 WRAP_RETURN(result);
633 struct cephwrap_pread_state {
634 ssize_t bytes_read;
635 struct vfs_aio_state vfs_aio_state;
639 * Fake up an async ceph read by calling the synchronous API.
641 static struct tevent_req *cephwrap_pread_send(struct vfs_handle_struct *handle,
642 TALLOC_CTX *mem_ctx,
643 struct tevent_context *ev,
644 struct files_struct *fsp,
645 void *data,
646 size_t n, off_t offset)
648 struct tevent_req *req = NULL;
649 struct cephwrap_pread_state *state = NULL;
650 int ret = -1;
652 DBG_DEBUG("[CEPH] %s\n", __func__);
653 req = tevent_req_create(mem_ctx, &state, struct cephwrap_pread_state);
654 if (req == NULL) {
655 return NULL;
658 ret = ceph_read(handle->data, fsp_get_io_fd(fsp), data, n, offset);
659 if (ret < 0) {
660 /* ceph returns -errno on error. */
661 tevent_req_error(req, -ret);
662 return tevent_req_post(req, ev);
665 state->bytes_read = ret;
666 tevent_req_done(req);
667 /* Return and schedule the completion of the call. */
668 return tevent_req_post(req, ev);
671 static ssize_t cephwrap_pread_recv(struct tevent_req *req,
672 struct vfs_aio_state *vfs_aio_state)
674 struct cephwrap_pread_state *state =
675 tevent_req_data(req, struct cephwrap_pread_state);
677 DBG_DEBUG("[CEPH] %s\n", __func__);
678 if (tevent_req_is_unix_error(req, &vfs_aio_state->error)) {
679 return -1;
681 *vfs_aio_state = state->vfs_aio_state;
682 return state->bytes_read;
685 static ssize_t cephwrap_pwrite(struct vfs_handle_struct *handle, files_struct *fsp, const void *data,
686 size_t n, off_t offset)
688 ssize_t result;
690 DBG_DEBUG("[CEPH] pwrite(%p, %p, %p, %llu, %llu)\n", handle, fsp, data, llu(n), llu(offset));
691 result = ceph_write(handle->data, fsp_get_io_fd(fsp), data, n, offset);
692 DBG_DEBUG("[CEPH] pwrite(...) = %llu\n", llu(result));
693 WRAP_RETURN(result);
696 struct cephwrap_pwrite_state {
697 ssize_t bytes_written;
698 struct vfs_aio_state vfs_aio_state;
702 * Fake up an async ceph write by calling the synchronous API.
704 static struct tevent_req *cephwrap_pwrite_send(struct vfs_handle_struct *handle,
705 TALLOC_CTX *mem_ctx,
706 struct tevent_context *ev,
707 struct files_struct *fsp,
708 const void *data,
709 size_t n, off_t offset)
711 struct tevent_req *req = NULL;
712 struct cephwrap_pwrite_state *state = NULL;
713 int ret = -1;
715 DBG_DEBUG("[CEPH] %s\n", __func__);
716 req = tevent_req_create(mem_ctx, &state, struct cephwrap_pwrite_state);
717 if (req == NULL) {
718 return NULL;
721 ret = ceph_write(handle->data, fsp_get_io_fd(fsp), data, n, offset);
722 if (ret < 0) {
723 /* ceph returns -errno on error. */
724 tevent_req_error(req, -ret);
725 return tevent_req_post(req, ev);
728 state->bytes_written = ret;
729 tevent_req_done(req);
730 /* Return and schedule the completion of the call. */
731 return tevent_req_post(req, ev);
734 static ssize_t cephwrap_pwrite_recv(struct tevent_req *req,
735 struct vfs_aio_state *vfs_aio_state)
737 struct cephwrap_pwrite_state *state =
738 tevent_req_data(req, struct cephwrap_pwrite_state);
740 DBG_DEBUG("[CEPH] %s\n", __func__);
741 if (tevent_req_is_unix_error(req, &vfs_aio_state->error)) {
742 return -1;
744 *vfs_aio_state = state->vfs_aio_state;
745 return state->bytes_written;
748 static off_t cephwrap_lseek(struct vfs_handle_struct *handle, files_struct *fsp, off_t offset, int whence)
750 off_t result = 0;
752 DBG_DEBUG("[CEPH] cephwrap_lseek\n");
753 result = ceph_lseek(handle->data, fsp_get_io_fd(fsp), offset, whence);
754 WRAP_RETURN(result);
757 static ssize_t cephwrap_sendfile(struct vfs_handle_struct *handle, int tofd, files_struct *fromfsp, const DATA_BLOB *hdr,
758 off_t offset, size_t n)
761 * We cannot support sendfile because libcephfs is in user space.
763 DBG_DEBUG("[CEPH] cephwrap_sendfile\n");
764 errno = ENOTSUP;
765 return -1;
768 static ssize_t cephwrap_recvfile(struct vfs_handle_struct *handle,
769 int fromfd,
770 files_struct *tofsp,
771 off_t offset,
772 size_t n)
775 * We cannot support recvfile because libcephfs is in user space.
777 DBG_DEBUG("[CEPH] cephwrap_recvfile\n");
778 errno=ENOTSUP;
779 return -1;
782 static int cephwrap_renameat(struct vfs_handle_struct *handle,
783 files_struct *srcfsp,
784 const struct smb_filename *smb_fname_src,
785 files_struct *dstfsp,
786 const struct smb_filename *smb_fname_dst)
788 struct smb_filename *full_fname_src = NULL;
789 struct smb_filename *full_fname_dst = NULL;
790 int result = -1;
792 DBG_DEBUG("[CEPH] cephwrap_renameat\n");
793 if (smb_fname_src->stream_name || smb_fname_dst->stream_name) {
794 errno = ENOENT;
795 return result;
798 full_fname_src = full_path_from_dirfsp_atname(talloc_tos(),
799 srcfsp,
800 smb_fname_src);
801 if (full_fname_src == NULL) {
802 errno = ENOMEM;
803 return -1;
805 full_fname_dst = full_path_from_dirfsp_atname(talloc_tos(),
806 dstfsp,
807 smb_fname_dst);
808 if (full_fname_dst == NULL) {
809 TALLOC_FREE(full_fname_src);
810 errno = ENOMEM;
811 return -1;
814 result = ceph_rename(handle->data,
815 full_fname_src->base_name,
816 full_fname_dst->base_name);
818 TALLOC_FREE(full_fname_src);
819 TALLOC_FREE(full_fname_dst);
821 WRAP_RETURN(result);
825 * Fake up an async ceph fsync by calling the synchronous API.
828 static struct tevent_req *cephwrap_fsync_send(struct vfs_handle_struct *handle,
829 TALLOC_CTX *mem_ctx,
830 struct tevent_context *ev,
831 files_struct *fsp)
833 struct tevent_req *req = NULL;
834 struct vfs_aio_state *state = NULL;
835 int ret = -1;
837 DBG_DEBUG("[CEPH] cephwrap_fsync_send\n");
839 req = tevent_req_create(mem_ctx, &state, struct vfs_aio_state);
840 if (req == NULL) {
841 return NULL;
844 /* Make sync call. */
845 ret = ceph_fsync(handle->data, fsp_get_io_fd(fsp), false);
847 if (ret != 0) {
848 /* ceph_fsync returns -errno on error. */
849 tevent_req_error(req, -ret);
850 return tevent_req_post(req, ev);
853 /* Mark it as done. */
854 tevent_req_done(req);
855 /* Return and schedule the completion of the call. */
856 return tevent_req_post(req, ev);
859 static int cephwrap_fsync_recv(struct tevent_req *req,
860 struct vfs_aio_state *vfs_aio_state)
862 struct vfs_aio_state *state =
863 tevent_req_data(req, struct vfs_aio_state);
865 DBG_DEBUG("[CEPH] cephwrap_fsync_recv\n");
867 if (tevent_req_is_unix_error(req, &vfs_aio_state->error)) {
868 return -1;
870 *vfs_aio_state = *state;
871 return 0;
874 #define SAMBA_STATX_ATTR_MASK (CEPH_STATX_BASIC_STATS|CEPH_STATX_BTIME)
876 static void init_stat_ex_from_ceph_statx(struct stat_ex *dst, const struct ceph_statx *stx)
878 DBG_DEBUG("[CEPH]\tstx = {dev = %llx, ino = %llu, mode = 0x%x, "
879 "nlink = %llu, uid = %d, gid = %d, rdev = %llx, size = %llu, "
880 "blksize = %llu, blocks = %llu, atime = %llu, mtime = %llu, "
881 "ctime = %llu, btime = %llu}\n",
882 llu(stx->stx_dev), llu(stx->stx_ino), stx->stx_mode,
883 llu(stx->stx_nlink), stx->stx_uid, stx->stx_gid,
884 llu(stx->stx_rdev), llu(stx->stx_size), llu(stx->stx_blksize),
885 llu(stx->stx_blocks), llu(stx->stx_atime.tv_sec),
886 llu(stx->stx_mtime.tv_sec), llu(stx->stx_ctime.tv_sec),
887 llu(stx->stx_btime.tv_sec));
889 if ((stx->stx_mask & SAMBA_STATX_ATTR_MASK) != SAMBA_STATX_ATTR_MASK) {
890 DBG_WARNING("%s: stx->stx_mask is incorrect (wanted %x, got %x)\n",
891 __func__, SAMBA_STATX_ATTR_MASK, stx->stx_mask);
894 dst->st_ex_dev = stx->stx_dev;
895 dst->st_ex_rdev = stx->stx_rdev;
896 dst->st_ex_ino = stx->stx_ino;
897 dst->st_ex_mode = stx->stx_mode;
898 dst->st_ex_uid = stx->stx_uid;
899 dst->st_ex_gid = stx->stx_gid;
900 dst->st_ex_size = stx->stx_size;
901 dst->st_ex_nlink = stx->stx_nlink;
902 dst->st_ex_atime = stx->stx_atime;
903 dst->st_ex_btime = stx->stx_btime;
904 dst->st_ex_ctime = stx->stx_ctime;
905 dst->st_ex_mtime = stx->stx_mtime;
906 dst->st_ex_blksize = stx->stx_blksize;
907 dst->st_ex_blocks = stx->stx_blocks;
910 static int cephwrap_stat(struct vfs_handle_struct *handle,
911 struct smb_filename *smb_fname)
913 int result = -1;
914 struct ceph_statx stx = { 0 };
916 DBG_DEBUG("[CEPH] stat(%p, %s)\n", handle, smb_fname_str_dbg(smb_fname));
918 if (smb_fname->stream_name) {
919 errno = ENOENT;
920 return result;
923 result = ceph_statx(handle->data, smb_fname->base_name, &stx,
924 SAMBA_STATX_ATTR_MASK, 0);
925 DBG_DEBUG("[CEPH] statx(...) = %d\n", result);
926 if (result < 0) {
927 WRAP_RETURN(result);
930 init_stat_ex_from_ceph_statx(&smb_fname->st, &stx);
931 DBG_DEBUG("[CEPH] mode = 0x%x\n", smb_fname->st.st_ex_mode);
932 return result;
935 static int cephwrap_fstat(struct vfs_handle_struct *handle, files_struct *fsp, SMB_STRUCT_STAT *sbuf)
937 int result = -1;
938 struct ceph_statx stx = { 0 };
939 int fd = fsp_get_pathref_fd(fsp);
941 DBG_DEBUG("[CEPH] fstat(%p, %d)\n", handle, fd);
942 result = ceph_fstatx(handle->data, fd, &stx,
943 SAMBA_STATX_ATTR_MASK, 0);
944 DBG_DEBUG("[CEPH] fstat(...) = %d\n", result);
945 if (result < 0) {
946 WRAP_RETURN(result);
949 init_stat_ex_from_ceph_statx(sbuf, &stx);
950 DBG_DEBUG("[CEPH] mode = 0x%x\n", sbuf->st_ex_mode);
951 return result;
954 static int cephwrap_fstatat(struct vfs_handle_struct *handle,
955 const struct files_struct *dirfsp,
956 const struct smb_filename *smb_fname,
957 SMB_STRUCT_STAT *sbuf,
958 int flags)
960 int result = -1;
961 struct ceph_statx stx = { 0 };
962 #ifdef HAVE_CEPH_STATXAT
963 int dirfd = fsp_get_pathref_fd(dirfsp);
965 DBG_DEBUG("[CEPH] fstatat(%p, %d, %s)\n",
966 handle, dirfd, smb_fname->base_name);
967 result = ceph_statxat(handle->data, dirfd, smb_fname->base_name,
968 &stx, SAMBA_STATX_ATTR_MASK, 0);
969 #else
970 struct smb_filename *full_fname = NULL;
972 full_fname = full_path_from_dirfsp_atname(talloc_tos(),
973 dirfsp,
974 smb_fname);
975 if (full_fname == NULL) {
976 errno = ENOMEM;
977 return -1;
980 DBG_DEBUG("[CEPH] fstatat(%p, %s)\n",
981 handle, smb_fname_str_dbg(full_fname));
982 result = ceph_statx(handle->data, full_fname->base_name,
983 &stx, SAMBA_STATX_ATTR_MASK, 0);
985 TALLOC_FREE(full_fname);
986 #endif
988 DBG_DEBUG("[CEPH] fstatat(...) = %d\n", result);
989 if (result < 0) {
990 WRAP_RETURN(result);
993 init_stat_ex_from_ceph_statx(sbuf, &stx);
994 DBG_DEBUG("[CEPH] mode = 0x%x\n", sbuf->st_ex_mode);
996 return 0;
999 static int cephwrap_lstat(struct vfs_handle_struct *handle,
1000 struct smb_filename *smb_fname)
1002 int result = -1;
1003 struct ceph_statx stx = { 0 };
1005 DBG_DEBUG("[CEPH] lstat(%p, %s)\n", handle, smb_fname_str_dbg(smb_fname));
1007 if (smb_fname->stream_name) {
1008 errno = ENOENT;
1009 return result;
1012 result = ceph_statx(handle->data, smb_fname->base_name, &stx,
1013 SAMBA_STATX_ATTR_MASK, AT_SYMLINK_NOFOLLOW);
1014 DBG_DEBUG("[CEPH] lstat(...) = %d\n", result);
1015 if (result < 0) {
1016 WRAP_RETURN(result);
1019 init_stat_ex_from_ceph_statx(&smb_fname->st, &stx);
1020 return result;
1023 static int cephwrap_fntimes(struct vfs_handle_struct *handle,
1024 files_struct *fsp,
1025 struct smb_file_time *ft)
1027 struct ceph_statx stx = { 0 };
1028 int result;
1029 int mask = 0;
1031 if (!is_omit_timespec(&ft->atime)) {
1032 stx.stx_atime = ft->atime;
1033 mask |= CEPH_SETATTR_ATIME;
1035 if (!is_omit_timespec(&ft->mtime)) {
1036 stx.stx_mtime = ft->mtime;
1037 mask |= CEPH_SETATTR_MTIME;
1039 if (!is_omit_timespec(&ft->create_time)) {
1040 stx.stx_btime = ft->create_time;
1041 mask |= CEPH_SETATTR_BTIME;
1044 if (!mask) {
1045 return 0;
1048 if (!fsp->fsp_flags.is_pathref) {
1050 * We can use an io_fd to set xattrs.
1052 result = ceph_fsetattrx(handle->data,
1053 fsp_get_io_fd(fsp),
1054 &stx,
1055 mask);
1056 } else {
1058 * This is no longer a handle based call.
1060 result = ceph_setattrx(handle->data,
1061 fsp->fsp_name->base_name,
1062 &stx,
1063 mask,
1067 DBG_DEBUG("[CEPH] ntimes(%p, %s, {%ld, %ld, %ld, %ld}) = %d\n",
1068 handle, fsp_str_dbg(fsp), ft->mtime.tv_sec, ft->atime.tv_sec,
1069 ft->ctime.tv_sec, ft->create_time.tv_sec, result);
1071 return result;
1074 static int cephwrap_unlinkat(struct vfs_handle_struct *handle,
1075 struct files_struct *dirfsp,
1076 const struct smb_filename *smb_fname,
1077 int flags)
1079 int result = -1;
1080 #ifdef HAVE_CEPH_UNLINKAT
1081 int dirfd = fsp_get_pathref_fd(dirfsp);
1083 DBG_DEBUG("[CEPH] unlinkat(%p, %d, %s)\n",
1084 handle,
1085 dirfd,
1086 smb_fname_str_dbg(smb_fname));
1088 if (smb_fname->stream_name) {
1089 errno = ENOENT;
1090 return result;
1093 result = ceph_unlinkat(handle->data,
1094 dirfd,
1095 smb_fname->base_name,
1096 flags);
1097 DBG_DEBUG("[CEPH] unlinkat(...) = %d\n", result);
1098 WRAP_RETURN(result);
1099 #else
1100 struct smb_filename *full_fname = NULL;
1102 DBG_DEBUG("[CEPH] unlink(%p, %s)\n",
1103 handle,
1104 smb_fname_str_dbg(smb_fname));
1106 if (smb_fname->stream_name) {
1107 errno = ENOENT;
1108 return result;
1111 full_fname = full_path_from_dirfsp_atname(talloc_tos(),
1112 dirfsp,
1113 smb_fname);
1114 if (full_fname == NULL) {
1115 return -1;
1118 if (flags & AT_REMOVEDIR) {
1119 result = ceph_rmdir(handle->data, full_fname->base_name);
1120 } else {
1121 result = ceph_unlink(handle->data, full_fname->base_name);
1123 TALLOC_FREE(full_fname);
1124 DBG_DEBUG("[CEPH] unlink(...) = %d\n", result);
1125 WRAP_RETURN(result);
1126 #endif
1129 static int cephwrap_fchmod(struct vfs_handle_struct *handle, files_struct *fsp, mode_t mode)
1131 int result;
1133 DBG_DEBUG("[CEPH] fchmod(%p, %p, %d)\n", handle, fsp, mode);
1134 if (!fsp->fsp_flags.is_pathref) {
1136 * We can use an io_fd to change permissions.
1138 result = ceph_fchmod(handle->data, fsp_get_io_fd(fsp), mode);
1139 } else {
1141 * This is no longer a handle based call.
1143 result = ceph_chmod(handle->data,
1144 fsp->fsp_name->base_name,
1145 mode);
1147 DBG_DEBUG("[CEPH] fchmod(...) = %d\n", result);
1148 WRAP_RETURN(result);
1151 static int cephwrap_fchown(struct vfs_handle_struct *handle, files_struct *fsp, uid_t uid, gid_t gid)
1153 int result;
1155 DBG_DEBUG("[CEPH] fchown(%p, %p, %d, %d)\n", handle, fsp, uid, gid);
1156 if (!fsp->fsp_flags.is_pathref) {
1158 * We can use an io_fd to change ownership.
1160 result = ceph_fchown(handle->data,
1161 fsp_get_io_fd(fsp),
1162 uid,
1163 gid);
1164 } else {
1166 * This is no longer a handle based call.
1168 result = ceph_chown(handle->data,
1169 fsp->fsp_name->base_name,
1170 uid,
1171 gid);
1174 DBG_DEBUG("[CEPH] fchown(...) = %d\n", result);
1175 WRAP_RETURN(result);
1178 static int cephwrap_lchown(struct vfs_handle_struct *handle,
1179 const struct smb_filename *smb_fname,
1180 uid_t uid,
1181 gid_t gid)
1183 int result;
1184 DBG_DEBUG("[CEPH] lchown(%p, %s, %d, %d)\n", handle, smb_fname->base_name, uid, gid);
1185 result = ceph_lchown(handle->data, smb_fname->base_name, uid, gid);
1186 DBG_DEBUG("[CEPH] lchown(...) = %d\n", result);
1187 WRAP_RETURN(result);
1190 static int cephwrap_chdir(struct vfs_handle_struct *handle,
1191 const struct smb_filename *smb_fname)
1193 int result = -1;
1194 DBG_DEBUG("[CEPH] chdir(%p, %s)\n", handle, smb_fname->base_name);
1195 result = ceph_chdir(handle->data, smb_fname->base_name);
1196 DBG_DEBUG("[CEPH] chdir(...) = %d\n", result);
1197 WRAP_RETURN(result);
1200 static struct smb_filename *cephwrap_getwd(struct vfs_handle_struct *handle,
1201 TALLOC_CTX *ctx)
1203 const char *cwd = ceph_getcwd(handle->data);
1204 DBG_DEBUG("[CEPH] getwd(%p) = %s\n", handle, cwd);
1205 return synthetic_smb_fname(ctx,
1206 cwd,
1207 NULL,
1208 NULL,
1213 static int strict_allocate_ftruncate(struct vfs_handle_struct *handle, files_struct *fsp, off_t len)
1215 off_t space_to_write;
1216 int result;
1217 NTSTATUS status;
1218 SMB_STRUCT_STAT *pst;
1220 status = vfs_stat_fsp(fsp);
1221 if (!NT_STATUS_IS_OK(status)) {
1222 return -1;
1224 pst = &fsp->fsp_name->st;
1226 #ifdef S_ISFIFO
1227 if (S_ISFIFO(pst->st_ex_mode))
1228 return 0;
1229 #endif
1231 if (pst->st_ex_size == len)
1232 return 0;
1234 /* Shrink - just ftruncate. */
1235 if (pst->st_ex_size > len) {
1236 result = ceph_ftruncate(handle->data, fsp_get_io_fd(fsp), len);
1237 WRAP_RETURN(result);
1240 space_to_write = len - pst->st_ex_size;
1241 result = ceph_fallocate(handle->data, fsp_get_io_fd(fsp), 0, pst->st_ex_size,
1242 space_to_write);
1243 WRAP_RETURN(result);
1246 static int cephwrap_ftruncate(struct vfs_handle_struct *handle, files_struct *fsp, off_t len)
1248 int result = -1;
1250 DBG_DEBUG("[CEPH] ftruncate(%p, %p, %llu\n", handle, fsp, llu(len));
1252 if (lp_strict_allocate(SNUM(fsp->conn))) {
1253 return strict_allocate_ftruncate(handle, fsp, len);
1256 result = ceph_ftruncate(handle->data, fsp_get_io_fd(fsp), len);
1257 WRAP_RETURN(result);
1260 static int cephwrap_fallocate(struct vfs_handle_struct *handle,
1261 struct files_struct *fsp,
1262 uint32_t mode,
1263 off_t offset,
1264 off_t len)
1266 int result;
1268 DBG_DEBUG("[CEPH] fallocate(%p, %p, %u, %llu, %llu\n",
1269 handle, fsp, mode, llu(offset), llu(len));
1270 /* unsupported mode flags are rejected by libcephfs */
1271 result = ceph_fallocate(handle->data, fsp_get_io_fd(fsp), mode, offset, len);
1272 DBG_DEBUG("[CEPH] fallocate(...) = %d\n", result);
1273 WRAP_RETURN(result);
1276 static bool cephwrap_lock(struct vfs_handle_struct *handle, files_struct *fsp, int op, off_t offset, off_t count, int type)
1278 DBG_DEBUG("[CEPH] lock\n");
1279 return true;
1282 static int cephwrap_filesystem_sharemode(struct vfs_handle_struct *handle,
1283 files_struct *fsp,
1284 uint32_t share_access,
1285 uint32_t access_mask)
1287 DBG_ERR("[CEPH] filesystem sharemodes unsupported! Consider setting "
1288 "\"kernel share modes = no\"\n");
1290 errno = ENOSYS;
1291 return -1;
1294 static int cephwrap_fcntl(vfs_handle_struct *handle,
1295 files_struct *fsp, int cmd, va_list cmd_arg)
1298 * SMB_VFS_FCNTL() is currently only called by vfs_set_blocking() to
1299 * clear O_NONBLOCK, etc for LOCK_MAND and FIFOs. Ignore it.
1301 if (cmd == F_GETFL) {
1302 return 0;
1303 } else if (cmd == F_SETFL) {
1304 va_list dup_cmd_arg;
1305 int opt;
1307 va_copy(dup_cmd_arg, cmd_arg);
1308 opt = va_arg(dup_cmd_arg, int);
1309 va_end(dup_cmd_arg);
1310 if (opt == 0) {
1311 return 0;
1313 DBG_ERR("unexpected fcntl SETFL(%d)\n", opt);
1314 goto err_out;
1316 DBG_ERR("unexpected fcntl: %d\n", cmd);
1317 err_out:
1318 errno = EINVAL;
1319 return -1;
1322 static bool cephwrap_getlock(struct vfs_handle_struct *handle, files_struct *fsp, off_t *poffset, off_t *pcount, int *ptype, pid_t *ppid)
1324 DBG_DEBUG("[CEPH] getlock returning false and errno=0\n");
1326 errno = 0;
1327 return false;
1331 * We cannot let this fall through to the default, because the file might only
1332 * be accessible from libcephfs (which is a user-space client) but the fd might
1333 * be for some file the kernel knows about.
1335 static int cephwrap_linux_setlease(struct vfs_handle_struct *handle, files_struct *fsp,
1336 int leasetype)
1338 int result = -1;
1340 DBG_DEBUG("[CEPH] linux_setlease\n");
1341 errno = ENOSYS;
1342 return result;
1345 static int cephwrap_symlinkat(struct vfs_handle_struct *handle,
1346 const struct smb_filename *link_target,
1347 struct files_struct *dirfsp,
1348 const struct smb_filename *new_smb_fname)
1350 int result = -1;
1351 #ifdef HAVE_CEPH_SYMLINKAT
1352 int dirfd = fsp_get_pathref_fd(dirfsp);
1354 DBG_DEBUG("[CEPH] symlinkat(%p, %s, %d, %s)\n",
1355 handle,
1356 link_target->base_name,
1357 dirfd,
1358 new_smb_fname->base_name);
1360 result = ceph_symlinkat(handle->data,
1361 link_target->base_name,
1362 dirfd,
1363 new_smb_fname->base_name);
1364 DBG_DEBUG("[CEPH] symlinkat(...) = %d\n", result);
1365 WRAP_RETURN(result);
1366 #else
1367 struct smb_filename *full_fname = NULL;
1369 full_fname = full_path_from_dirfsp_atname(talloc_tos(),
1370 dirfsp,
1371 new_smb_fname);
1372 if (full_fname == NULL) {
1373 return -1;
1376 DBG_DEBUG("[CEPH] symlink(%p, %s, %s)\n", handle,
1377 link_target->base_name,
1378 full_fname->base_name);
1380 result = ceph_symlink(handle->data,
1381 link_target->base_name,
1382 full_fname->base_name);
1383 TALLOC_FREE(full_fname);
1384 DBG_DEBUG("[CEPH] symlink(...) = %d\n", result);
1385 WRAP_RETURN(result);
1386 #endif
1389 static int cephwrap_readlinkat(struct vfs_handle_struct *handle,
1390 const struct files_struct *dirfsp,
1391 const struct smb_filename *smb_fname,
1392 char *buf,
1393 size_t bufsiz)
1395 int result = -1;
1396 #ifdef HAVE_CEPH_READLINKAT
1397 int dirfd = fsp_get_pathref_fd(dirfsp);
1399 DBG_DEBUG("[CEPH] readlinkat(%p, %d, %s, %p, %llu)\n",
1400 handle,
1401 dirfd,
1402 smb_fname->base_name,
1403 buf,
1404 llu(bufsiz));
1406 result = ceph_readlinkat(handle->data,
1407 dirfd,
1408 smb_fname->base_name,
1409 buf,
1410 bufsiz);
1412 DBG_DEBUG("[CEPH] readlinkat(...) = %d\n", result);
1413 WRAP_RETURN(result);
1414 #else
1415 struct smb_filename *full_fname = NULL;
1417 full_fname = full_path_from_dirfsp_atname(talloc_tos(),
1418 dirfsp,
1419 smb_fname);
1420 if (full_fname == NULL) {
1421 return -1;
1424 DBG_DEBUG("[CEPH] readlink(%p, %s, %p, %llu)\n", handle,
1425 full_fname->base_name, buf, llu(bufsiz));
1427 result = ceph_readlink(handle->data, full_fname->base_name, buf, bufsiz);
1428 TALLOC_FREE(full_fname);
1429 DBG_DEBUG("[CEPH] readlink(...) = %d\n", result);
1430 WRAP_RETURN(result);
1431 #endif
1434 static int cephwrap_linkat(struct vfs_handle_struct *handle,
1435 files_struct *srcfsp,
1436 const struct smb_filename *old_smb_fname,
1437 files_struct *dstfsp,
1438 const struct smb_filename *new_smb_fname,
1439 int flags)
1441 struct smb_filename *full_fname_old = NULL;
1442 struct smb_filename *full_fname_new = NULL;
1443 int result = -1;
1445 full_fname_old = full_path_from_dirfsp_atname(talloc_tos(),
1446 srcfsp,
1447 old_smb_fname);
1448 if (full_fname_old == NULL) {
1449 return -1;
1451 full_fname_new = full_path_from_dirfsp_atname(talloc_tos(),
1452 dstfsp,
1453 new_smb_fname);
1454 if (full_fname_new == NULL) {
1455 TALLOC_FREE(full_fname_old);
1456 return -1;
1459 DBG_DEBUG("[CEPH] link(%p, %s, %s)\n", handle,
1460 full_fname_old->base_name,
1461 full_fname_new->base_name);
1463 result = ceph_link(handle->data,
1464 full_fname_old->base_name,
1465 full_fname_new->base_name);
1466 DBG_DEBUG("[CEPH] link(...) = %d\n", result);
1467 TALLOC_FREE(full_fname_old);
1468 TALLOC_FREE(full_fname_new);
1469 WRAP_RETURN(result);
1472 static int cephwrap_mknodat(struct vfs_handle_struct *handle,
1473 files_struct *dirfsp,
1474 const struct smb_filename *smb_fname,
1475 mode_t mode,
1476 SMB_DEV_T dev)
1478 struct smb_filename *full_fname = NULL;
1479 int result = -1;
1481 full_fname = full_path_from_dirfsp_atname(talloc_tos(),
1482 dirfsp,
1483 smb_fname);
1484 if (full_fname == NULL) {
1485 return -1;
1488 DBG_DEBUG("[CEPH] mknodat(%p, %s)\n", handle, full_fname->base_name);
1489 result = ceph_mknod(handle->data, full_fname->base_name, mode, dev);
1490 DBG_DEBUG("[CEPH] mknodat(...) = %d\n", result);
1492 TALLOC_FREE(full_fname);
1494 WRAP_RETURN(result);
1498 * This is a simple version of real-path ... a better version is needed to
1499 * ask libcephfs about symbolic links.
1501 static struct smb_filename *cephwrap_realpath(struct vfs_handle_struct *handle,
1502 TALLOC_CTX *ctx,
1503 const struct smb_filename *smb_fname)
1505 char *result = NULL;
1506 const char *path = smb_fname->base_name;
1507 size_t len = strlen(path);
1508 struct smb_filename *result_fname = NULL;
1509 int r = -1;
1511 if (len && (path[0] == '/')) {
1512 r = asprintf(&result, "%s", path);
1513 } else if ((len >= 2) && (path[0] == '.') && (path[1] == '/')) {
1514 if (len == 2) {
1515 r = asprintf(&result, "%s",
1516 handle->conn->cwd_fsp->fsp_name->base_name);
1517 } else {
1518 r = asprintf(&result, "%s/%s",
1519 handle->conn->cwd_fsp->fsp_name->base_name, &path[2]);
1521 } else {
1522 r = asprintf(&result, "%s/%s",
1523 handle->conn->cwd_fsp->fsp_name->base_name, path);
1526 if (r < 0) {
1527 return NULL;
1530 DBG_DEBUG("[CEPH] realpath(%p, %s) = %s\n", handle, path, result);
1531 result_fname = synthetic_smb_fname(ctx,
1532 result,
1533 NULL,
1534 NULL,
1537 SAFE_FREE(result);
1538 return result_fname;
1542 static int cephwrap_fchflags(struct vfs_handle_struct *handle,
1543 struct files_struct *fsp,
1544 unsigned int flags)
1546 errno = ENOSYS;
1547 return -1;
1550 static NTSTATUS cephwrap_get_real_filename_at(
1551 struct vfs_handle_struct *handle,
1552 struct files_struct *dirfsp,
1553 const char *name,
1554 TALLOC_CTX *mem_ctx,
1555 char **found_name)
1558 * Don't fall back to get_real_filename so callers can differentiate
1559 * between a full directory scan and an actual case-insensitive stat.
1561 return NT_STATUS_NOT_SUPPORTED;
1564 static const char *cephwrap_connectpath(
1565 struct vfs_handle_struct *handle,
1566 const struct files_struct *dirfsp,
1567 const struct smb_filename *smb_fname)
1569 return handle->conn->connectpath;
1572 static NTSTATUS cephwrap_fget_dos_attributes(struct vfs_handle_struct *handle,
1573 struct files_struct *fsp,
1574 uint32_t *dosmode)
1576 struct timespec saved_btime = fsp->fsp_name->st.st_ex_btime;
1577 NTSTATUS status;
1579 status = fget_ea_dos_attribute(fsp, dosmode);
1582 * Restore previously stored btime from statx timestamps as it should be
1583 * the only source of truth. create_time from dos attribute, if any, may
1584 * have older values which isn't trustworthy to be looked at for other
1585 * open file handle operations.
1587 fsp->fsp_name->st.st_ex_btime = saved_btime;
1589 return status;
1592 static NTSTATUS cephwrap_fset_dos_attributes(struct vfs_handle_struct *handle,
1593 struct files_struct *fsp,
1594 uint32_t dosmode)
1596 struct timespec saved_btime = fsp->fsp_name->st.st_ex_btime;
1597 NTSTATUS status;
1599 status = set_ea_dos_attribute(handle->conn, fsp->fsp_name, dosmode);
1602 * Restore previously stored btime from statx timestamps. This is done
1603 * to ensure that we have the exact btime in fsp stat information while
1604 * the file handle is still open since the create_time stored as part of
1605 * dos attributes can loose its precision when converted back to btime.
1607 fsp->fsp_name->st.st_ex_btime = saved_btime;
1609 return status;
1612 /****************************************************************
1613 Extended attribute operations.
1614 *****************************************************************/
1616 static ssize_t cephwrap_fgetxattr(struct vfs_handle_struct *handle,
1617 struct files_struct *fsp,
1618 const char *name,
1619 void *value,
1620 size_t size)
1622 int ret;
1623 DBG_DEBUG("[CEPH] fgetxattr(%p, %p, %s, %p, %llu)\n",
1624 handle,
1625 fsp,
1626 name,
1627 value,
1628 llu(size));
1629 if (!fsp->fsp_flags.is_pathref) {
1630 ret = ceph_fgetxattr(handle->data,
1631 fsp_get_io_fd(fsp),
1632 name,
1633 value,
1634 size);
1635 } else {
1636 ret = ceph_getxattr(handle->data,
1637 fsp->fsp_name->base_name,
1638 name,
1639 value,
1640 size);
1642 DBG_DEBUG("[CEPH] fgetxattr(...) = %d\n", ret);
1643 if (ret < 0) {
1644 WRAP_RETURN(ret);
1646 return (ssize_t)ret;
1649 static ssize_t cephwrap_flistxattr(struct vfs_handle_struct *handle, struct files_struct *fsp, char *list, size_t size)
1651 int ret;
1652 DBG_DEBUG("[CEPH] flistxattr(%p, %p, %p, %llu)\n",
1653 handle, fsp, list, llu(size));
1654 if (!fsp->fsp_flags.is_pathref) {
1656 * We can use an io_fd to list xattrs.
1658 ret = ceph_flistxattr(handle->data,
1659 fsp_get_io_fd(fsp),
1660 list,
1661 size);
1662 } else {
1664 * This is no longer a handle based call.
1666 ret = ceph_listxattr(handle->data,
1667 fsp->fsp_name->base_name,
1668 list,
1669 size);
1671 DBG_DEBUG("[CEPH] flistxattr(...) = %d\n", ret);
1672 if (ret < 0) {
1673 WRAP_RETURN(ret);
1675 return (ssize_t)ret;
1678 static int cephwrap_fremovexattr(struct vfs_handle_struct *handle, struct files_struct *fsp, const char *name)
1680 int ret;
1681 DBG_DEBUG("[CEPH] fremovexattr(%p, %p, %s)\n", handle, fsp, name);
1682 if (!fsp->fsp_flags.is_pathref) {
1684 * We can use an io_fd to remove xattrs.
1686 ret = ceph_fremovexattr(handle->data, fsp_get_io_fd(fsp), name);
1687 } else {
1689 * This is no longer a handle based call.
1691 ret = ceph_removexattr(handle->data,
1692 fsp->fsp_name->base_name,
1693 name);
1695 DBG_DEBUG("[CEPH] fremovexattr(...) = %d\n", ret);
1696 WRAP_RETURN(ret);
1699 static int cephwrap_fsetxattr(struct vfs_handle_struct *handle, struct files_struct *fsp, const char *name, const void *value, size_t size, int flags)
1701 int ret;
1702 DBG_DEBUG("[CEPH] fsetxattr(%p, %p, %s, %p, %llu, %d)\n", handle, fsp, name, value, llu(size), flags);
1703 if (!fsp->fsp_flags.is_pathref) {
1705 * We can use an io_fd to set xattrs.
1707 ret = ceph_fsetxattr(handle->data,
1708 fsp_get_io_fd(fsp),
1709 name,
1710 value,
1711 size,
1712 flags);
1713 } else {
1715 * This is no longer a handle based call.
1717 ret = ceph_setxattr(handle->data,
1718 fsp->fsp_name->base_name,
1719 name,
1720 value,
1721 size,
1722 flags);
1724 DBG_DEBUG("[CEPH] fsetxattr(...) = %d\n", ret);
1725 WRAP_RETURN(ret);
1728 static bool cephwrap_aio_force(struct vfs_handle_struct *handle, struct files_struct *fsp)
1732 * We do not support AIO yet.
1735 DBG_DEBUG("[CEPH] cephwrap_aio_force(%p, %p) = false (errno = ENOTSUP)\n", handle, fsp);
1736 errno = ENOTSUP;
1737 return false;
1740 static NTSTATUS cephwrap_create_dfs_pathat(struct vfs_handle_struct *handle,
1741 struct files_struct *dirfsp,
1742 const struct smb_filename *smb_fname,
1743 const struct referral *reflist,
1744 size_t referral_count)
1746 TALLOC_CTX *frame = talloc_stackframe();
1747 NTSTATUS status = NT_STATUS_NO_MEMORY;
1748 int ret;
1749 char *msdfs_link = NULL;
1750 struct smb_filename *full_fname = NULL;
1752 full_fname = full_path_from_dirfsp_atname(talloc_tos(),
1753 dirfsp,
1754 smb_fname);
1755 if (full_fname == NULL) {
1756 goto out;
1759 /* Form the msdfs_link contents */
1760 msdfs_link = msdfs_link_string(frame,
1761 reflist,
1762 referral_count);
1763 if (msdfs_link == NULL) {
1764 goto out;
1767 ret = ceph_symlink(handle->data,
1768 msdfs_link,
1769 full_fname->base_name);
1770 if (ret == 0) {
1771 status = NT_STATUS_OK;
1772 } else {
1773 status = map_nt_error_from_unix(-ret);
1776 out:
1778 DBG_DEBUG("[CEPH] create_dfs_pathat(%s) = %s\n",
1779 full_fname != NULL ? full_fname->base_name : "",
1780 nt_errstr(status));
1782 TALLOC_FREE(frame);
1783 return status;
1787 * Read and return the contents of a DFS redirect given a
1788 * pathname. A caller can pass in NULL for ppreflist and
1789 * preferral_count but still determine if this was a
1790 * DFS redirect point by getting NT_STATUS_OK back
1791 * without incurring the overhead of reading and parsing
1792 * the referral contents.
1795 static NTSTATUS cephwrap_read_dfs_pathat(struct vfs_handle_struct *handle,
1796 TALLOC_CTX *mem_ctx,
1797 struct files_struct *dirfsp,
1798 struct smb_filename *smb_fname,
1799 struct referral **ppreflist,
1800 size_t *preferral_count)
1802 NTSTATUS status = NT_STATUS_NO_MEMORY;
1803 size_t bufsize;
1804 char *link_target = NULL;
1805 int referral_len;
1806 bool ok;
1807 #if defined(HAVE_BROKEN_READLINK)
1808 char link_target_buf[PATH_MAX];
1809 #else
1810 char link_target_buf[7];
1811 #endif
1812 struct ceph_statx stx = { 0 };
1813 struct smb_filename *full_fname = NULL;
1814 int ret;
1816 if (is_named_stream(smb_fname)) {
1817 status = NT_STATUS_OBJECT_NAME_NOT_FOUND;
1818 goto err;
1821 if (ppreflist == NULL && preferral_count == NULL) {
1823 * We're only checking if this is a DFS
1824 * redirect. We don't need to return data.
1826 bufsize = sizeof(link_target_buf);
1827 link_target = link_target_buf;
1828 } else {
1829 bufsize = PATH_MAX;
1830 link_target = talloc_array(mem_ctx, char, bufsize);
1831 if (!link_target) {
1832 goto err;
1836 full_fname = full_path_from_dirfsp_atname(talloc_tos(),
1837 dirfsp,
1838 smb_fname);
1839 if (full_fname == NULL) {
1840 status = NT_STATUS_NO_MEMORY;
1841 goto err;
1844 ret = ceph_statx(handle->data,
1845 full_fname->base_name,
1846 &stx,
1847 SAMBA_STATX_ATTR_MASK,
1848 AT_SYMLINK_NOFOLLOW);
1849 if (ret < 0) {
1850 status = map_nt_error_from_unix(-ret);
1851 goto err;
1854 referral_len = ceph_readlink(handle->data,
1855 full_fname->base_name,
1856 link_target,
1857 bufsize - 1);
1858 if (referral_len < 0) {
1859 /* ceph errors are -errno. */
1860 if (-referral_len == EINVAL) {
1861 DBG_INFO("%s is not a link.\n",
1862 full_fname->base_name);
1863 status = NT_STATUS_OBJECT_TYPE_MISMATCH;
1864 } else {
1865 status = map_nt_error_from_unix(-referral_len);
1866 DBG_ERR("Error reading "
1867 "msdfs link %s: %s\n",
1868 full_fname->base_name,
1869 strerror(errno));
1871 goto err;
1873 link_target[referral_len] = '\0';
1875 DBG_INFO("%s -> %s\n",
1876 full_fname->base_name,
1877 link_target);
1879 if (!strnequal(link_target, "msdfs:", 6)) {
1880 status = NT_STATUS_OBJECT_TYPE_MISMATCH;
1881 goto err;
1884 if (ppreflist == NULL && preferral_count == NULL) {
1885 /* Early return for checking if this is a DFS link. */
1886 TALLOC_FREE(full_fname);
1887 init_stat_ex_from_ceph_statx(&smb_fname->st, &stx);
1888 return NT_STATUS_OK;
1891 ok = parse_msdfs_symlink(mem_ctx,
1892 lp_msdfs_shuffle_referrals(SNUM(handle->conn)),
1893 link_target,
1894 ppreflist,
1895 preferral_count);
1897 if (ok) {
1898 init_stat_ex_from_ceph_statx(&smb_fname->st, &stx);
1899 status = NT_STATUS_OK;
1900 } else {
1901 status = NT_STATUS_NO_MEMORY;
1904 err:
1906 if (link_target != link_target_buf) {
1907 TALLOC_FREE(link_target);
1909 TALLOC_FREE(full_fname);
1910 return status;
1913 static struct vfs_fn_pointers ceph_fns = {
1914 /* Disk operations */
1916 .connect_fn = cephwrap_connect,
1917 .disconnect_fn = cephwrap_disconnect,
1918 .disk_free_fn = cephwrap_disk_free,
1919 .get_quota_fn = cephwrap_get_quota,
1920 .set_quota_fn = cephwrap_set_quota,
1921 .statvfs_fn = cephwrap_statvfs,
1922 .fs_capabilities_fn = cephwrap_fs_capabilities,
1924 /* Directory operations */
1926 .fdopendir_fn = cephwrap_fdopendir,
1927 .readdir_fn = cephwrap_readdir,
1928 .rewind_dir_fn = cephwrap_rewinddir,
1929 .mkdirat_fn = cephwrap_mkdirat,
1930 .closedir_fn = cephwrap_closedir,
1932 /* File operations */
1934 .create_dfs_pathat_fn = cephwrap_create_dfs_pathat,
1935 .read_dfs_pathat_fn = cephwrap_read_dfs_pathat,
1936 .openat_fn = cephwrap_openat,
1937 .close_fn = cephwrap_close,
1938 .pread_fn = cephwrap_pread,
1939 .pread_send_fn = cephwrap_pread_send,
1940 .pread_recv_fn = cephwrap_pread_recv,
1941 .pwrite_fn = cephwrap_pwrite,
1942 .pwrite_send_fn = cephwrap_pwrite_send,
1943 .pwrite_recv_fn = cephwrap_pwrite_recv,
1944 .lseek_fn = cephwrap_lseek,
1945 .sendfile_fn = cephwrap_sendfile,
1946 .recvfile_fn = cephwrap_recvfile,
1947 .renameat_fn = cephwrap_renameat,
1948 .fsync_send_fn = cephwrap_fsync_send,
1949 .fsync_recv_fn = cephwrap_fsync_recv,
1950 .stat_fn = cephwrap_stat,
1951 .fstat_fn = cephwrap_fstat,
1952 .lstat_fn = cephwrap_lstat,
1953 .fstatat_fn = cephwrap_fstatat,
1954 .unlinkat_fn = cephwrap_unlinkat,
1955 .fchmod_fn = cephwrap_fchmod,
1956 .fchown_fn = cephwrap_fchown,
1957 .lchown_fn = cephwrap_lchown,
1958 .chdir_fn = cephwrap_chdir,
1959 .getwd_fn = cephwrap_getwd,
1960 .fntimes_fn = cephwrap_fntimes,
1961 .ftruncate_fn = cephwrap_ftruncate,
1962 .fallocate_fn = cephwrap_fallocate,
1963 .lock_fn = cephwrap_lock,
1964 .filesystem_sharemode_fn = cephwrap_filesystem_sharemode,
1965 .fcntl_fn = cephwrap_fcntl,
1966 .linux_setlease_fn = cephwrap_linux_setlease,
1967 .getlock_fn = cephwrap_getlock,
1968 .symlinkat_fn = cephwrap_symlinkat,
1969 .readlinkat_fn = cephwrap_readlinkat,
1970 .linkat_fn = cephwrap_linkat,
1971 .mknodat_fn = cephwrap_mknodat,
1972 .realpath_fn = cephwrap_realpath,
1973 .fchflags_fn = cephwrap_fchflags,
1974 .get_real_filename_at_fn = cephwrap_get_real_filename_at,
1975 .connectpath_fn = cephwrap_connectpath,
1976 .fget_dos_attributes_fn = cephwrap_fget_dos_attributes,
1977 .fset_dos_attributes_fn = cephwrap_fset_dos_attributes,
1979 /* EA operations. */
1980 .getxattrat_send_fn = vfs_not_implemented_getxattrat_send,
1981 .getxattrat_recv_fn = vfs_not_implemented_getxattrat_recv,
1982 .fgetxattr_fn = cephwrap_fgetxattr,
1983 .flistxattr_fn = cephwrap_flistxattr,
1984 .fremovexattr_fn = cephwrap_fremovexattr,
1985 .fsetxattr_fn = cephwrap_fsetxattr,
1987 /* Posix ACL Operations */
1988 .sys_acl_get_fd_fn = posixacl_xattr_acl_get_fd,
1989 .sys_acl_blob_get_fd_fn = posix_sys_acl_blob_get_fd,
1990 .sys_acl_set_fd_fn = posixacl_xattr_acl_set_fd,
1991 .sys_acl_delete_def_fd_fn = posixacl_xattr_acl_delete_def_fd,
1993 /* aio operations */
1994 .aio_force_fn = cephwrap_aio_force,
1997 static_decl_vfs;
1998 NTSTATUS vfs_ceph_init(TALLOC_CTX *ctx)
2000 return smb_register_vfs(SMB_VFS_INTERFACE_VERSION,
2001 "ceph", &ceph_fns);