debug: Set close-on-exec for the main log file FD
[Samba.git] / source3 / modules / vfs_ceph.c
blobb0a00249f16e0a48d9a47cd13fbeddcc55028ea3
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 libceph.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 = ceph [any others you need go here]
33 #include "includes.h"
34 #include "smbd/smbd.h"
35 #include <dirent.h>
36 #include <sys/statvfs.h>
37 #include "cephfs/libcephfs.h"
38 #include "smbprofile.h"
40 #undef DBGC_CLASS
41 #define DBGC_CLASS DBGC_VFS
44 * Use %llu whenever we have a 64bit unsigned int, and cast to (long long unsigned)
46 #define llu(_var) ((long long unsigned)_var)
49 * Note, libceph's return code model is to return -errno! So we have to convert
50 * to what Samba expects, with is set errno to -return and return -1
52 #define WRAP_RETURN(_res) \
53 errno = 0; \
54 if (_res < 0) { \
55 errno = -_res; \
56 return -1; \
57 } \
58 return _res \
61 * We mount only one file system and then all shares are assumed to be in that.
62 * FIXME: If we want to support more than one FS, then we have to deal with
63 * this differently.
65 * So, cmount tells us if we have been this way before and whether
66 * we need to mount ceph and cmount_cnt tells us how many times we have
67 * connected
69 static struct ceph_mount_info * cmount = NULL;
70 static uint32_t cmount_cnt = 0;
72 /* Check for NULL pointer parameters in cephwrap_* functions */
74 /* We don't want to have NULL function pointers lying around. Someone
75 is sure to try and execute them. These stubs are used to prevent
76 this possibility. */
78 static int cephwrap_connect(struct vfs_handle_struct *handle, const char *service, const char *user)
80 int ret;
81 char buf[256];
83 const char * conf_file;
85 if (cmount) {
86 handle->data = cmount; /* We have been here before */
87 cmount_cnt++;
88 return 0;
91 conf_file = lp_parm_const_string(SNUM(handle->conn), "ceph", "config_file", NULL);
93 DEBUG(2, ( "[CEPH] calling: ceph_create\n" ));
94 ret = ceph_create(&cmount, NULL);
95 if (ret)
96 goto err_out;
98 if (conf_file) {
99 /* Override the config file */
100 DEBUG(2, ( "[CEPH] calling: ceph_conf_read_file\n" ));
101 ret = ceph_conf_read_file(cmount, conf_file);
102 } else {
104 DEBUG(2, ( "[CEPH] calling: ceph_conf_read_file with %s\n", conf_file));
105 ret = ceph_conf_read_file(cmount, NULL);
108 if (ret)
109 goto err_out;
111 DEBUG(2, ( "[CEPH] calling: ceph_conf_get\n" ));
112 ret = ceph_conf_get(cmount, "log file", buf, sizeof(buf));
113 if (ret < 0)
114 goto err_out;
116 DEBUG(2, ("[CEPH] calling: ceph_mount\n"));
117 ret = ceph_mount(cmount, NULL);
118 if (ret < 0)
119 goto err_out;
123 * encode mount context/state into our vfs/connection holding structure
124 * cmount is a ceph_mount_t*
126 handle->data = cmount;
127 cmount_cnt++;
129 return 0;
131 err_out:
133 * Handle the error correctly. Ceph returns -errno.
135 DEBUG(2, ("[CEPH] Error return: %s\n", strerror(-ret)));
136 WRAP_RETURN(ret);
139 static void cephwrap_disconnect(struct vfs_handle_struct *handle)
141 if (!cmount) {
142 DEBUG(0, ("[CEPH] Error, ceph not mounted\n"));
143 return;
146 /* Should we unmount/shutdown? Only if the last disconnect? */
147 if (--cmount_cnt) {
148 DEBUG(10, ("[CEPH] Not shuting down CEPH because still more connections\n"));
149 return;
152 ceph_shutdown(cmount);
154 cmount = NULL; /* Make it safe */
157 /* Disk operations */
159 static uint64_t cephwrap_disk_free(struct vfs_handle_struct *handle, const char *path, bool small_query, uint64_t *bsize,
160 uint64_t *dfree, uint64_t *dsize)
162 struct statvfs statvfs_buf;
163 int ret;
165 if (!(ret = ceph_statfs(handle->data, path, &statvfs_buf))) {
167 * Provide all the correct values.
169 *bsize = statvfs_buf.f_bsize;
170 *dfree = statvfs_buf.f_bsize * statvfs_buf.f_bavail;
171 *dsize = statvfs_buf.f_bsize * statvfs_buf.f_blocks;
172 DEBUG(10, ("[CEPH] bsize: %llu, dfree: %llu, dsize: %llu\n",
173 llu(*bsize), llu(*dfree), llu(*dsize)));
174 return *dfree;
175 } else {
176 DEBUG(10, ("[CEPH] ceph_statfs returned %d\n", ret));
177 WRAP_RETURN(ret);
181 static int cephwrap_get_quota(struct vfs_handle_struct *handle, enum SMB_QUOTA_TYPE qtype, unid_t id, SMB_DISK_QUOTA *qt)
183 /* libceph: Ceph does not implement this */
184 #if 0
185 /* was ifdef HAVE_SYS_QUOTAS */
186 int ret;
188 ret = ceph_get_quota(handle->conn->connectpath, qtype, id, qt);
190 if (ret) {
191 errno = -ret;
192 ret = -1;
195 return ret;
196 #else
197 errno = ENOSYS;
198 return -1;
199 #endif
202 static int cephwrap_set_quota(struct vfs_handle_struct *handle, enum SMB_QUOTA_TYPE qtype, unid_t id, SMB_DISK_QUOTA *qt)
204 /* libceph: Ceph does not implement this */
205 #if 0
206 /* was ifdef HAVE_SYS_QUOTAS */
207 int ret;
209 ret = ceph_set_quota(handle->conn->connectpath, qtype, id, qt);
210 if (ret) {
211 errno = -ret;
212 ret = -1;
215 return ret;
216 #else
217 WRAP_RETURN(-ENOSYS);
218 #endif
221 static int cephwrap_statvfs(struct vfs_handle_struct *handle, const char *path, vfs_statvfs_struct *statbuf)
223 struct statvfs statvfs_buf;
224 int ret;
226 ret = ceph_statfs(handle->data, path, &statvfs_buf);
227 if (ret < 0) {
228 WRAP_RETURN(ret);
229 } else {
230 statbuf->OptimalTransferSize = statvfs_buf.f_frsize;
231 statbuf->BlockSize = statvfs_buf.f_bsize;
232 statbuf->TotalBlocks = statvfs_buf.f_blocks;
233 statbuf->BlocksAvail = statvfs_buf.f_bfree;
234 statbuf->UserBlocksAvail = statvfs_buf.f_bavail;
235 statbuf->TotalFileNodes = statvfs_buf.f_files;
236 statbuf->FreeFileNodes = statvfs_buf.f_ffree;
237 statbuf->FsIdentifier = statvfs_buf.f_fsid;
238 DEBUG(10, ("[CEPH] f_bsize: %ld, f_blocks: %ld, f_bfree: %ld, f_bavail: %ld\n",
239 (long int)statvfs_buf.f_bsize, (long int)statvfs_buf.f_blocks,
240 (long int)statvfs_buf.f_bfree, (long int)statvfs_buf.f_bavail));
242 return ret;
245 /* Directory operations */
247 static DIR *cephwrap_opendir(struct vfs_handle_struct *handle, const char *fname, const char *mask, uint32 attr)
249 int ret = 0;
250 struct ceph_dir_result *result;
251 DEBUG(10, ("[CEPH] opendir(%p, %s)\n", handle, fname));
253 /* Returns NULL if it does not exist or there are problems ? */
254 ret = ceph_opendir(handle->data, fname, &result);
255 if (ret < 0) {
256 result = NULL;
257 errno = -ret; /* We return result which is NULL in this case */
260 DEBUG(10, ("[CEPH] opendir(...) = %d\n", ret));
261 return (DIR *) result;
264 static DIR *cephwrap_fdopendir(struct vfs_handle_struct *handle,
265 struct files_struct *fsp,
266 const char *mask,
267 uint32 attributes)
269 int ret = 0;
270 struct ceph_dir_result *result;
271 DEBUG(10, ("[CEPH] fdopendir(%p, %p)\n", handle, fsp));
273 ret = ceph_opendir(handle->data, fsp->fsp_name->base_name, &result);
274 if (ret < 0) {
275 result = NULL;
276 errno = -ret; /* We return result which is NULL in this case */
279 DEBUG(10, ("[CEPH] fdopendir(...) = %d\n", ret));
280 return (DIR *) result;
283 static struct dirent *cephwrap_readdir(struct vfs_handle_struct *handle,
284 DIR *dirp,
285 SMB_STRUCT_STAT *sbuf)
287 struct dirent *result;
289 DEBUG(10, ("[CEPH] readdir(%p, %p)\n", handle, dirp));
290 result = ceph_readdir(handle->data, (struct ceph_dir_result *) dirp);
291 DEBUG(10, ("[CEPH] readdir(...) = %p\n", result));
293 /* Default Posix readdir() does not give us stat info.
294 * Set to invalid to indicate we didn't return this info. */
295 if (sbuf)
296 SET_STAT_INVALID(*sbuf);
297 return result;
300 static void cephwrap_seekdir(struct vfs_handle_struct *handle, DIR *dirp, long offset)
302 DEBUG(10, ("[CEPH] seekdir(%p, %p, %ld)\n", handle, dirp, offset));
303 ceph_seekdir(handle->data, (struct ceph_dir_result *) dirp, offset);
306 static long cephwrap_telldir(struct vfs_handle_struct *handle, DIR *dirp)
308 long ret;
309 DEBUG(10, ("[CEPH] telldir(%p, %p)\n", handle, dirp));
310 ret = ceph_telldir(handle->data, (struct ceph_dir_result *) dirp);
311 DEBUG(10, ("[CEPH] telldir(...) = %ld\n", ret));
312 WRAP_RETURN(ret);
315 static void cephwrap_rewinddir(struct vfs_handle_struct *handle, DIR *dirp)
317 DEBUG(10, ("[CEPH] rewinddir(%p, %p)\n", handle, dirp));
318 ceph_rewinddir(handle->data, (struct ceph_dir_result *) dirp);
321 static int cephwrap_mkdir(struct vfs_handle_struct *handle, const char *path, mode_t mode)
323 int result;
324 bool has_dacl = False;
325 char *parent = NULL;
327 DEBUG(10, ("[CEPH] mkdir(%p, %s)\n", handle, path));
329 if (lp_inherit_acls(SNUM(handle->conn))
330 && parent_dirname(talloc_tos(), path, &parent, NULL)
331 && (has_dacl = directory_has_default_acl(handle->conn, parent)))
332 mode = 0777;
334 TALLOC_FREE(parent);
336 result = ceph_mkdir(handle->data, path, mode);
339 * Note. This order is important
341 if (result) {
342 WRAP_RETURN(result);
343 } else if (result == 0 && !has_dacl) {
345 * We need to do this as the default behavior of POSIX ACLs
346 * is to set the mask to be the requested group permission
347 * bits, not the group permission bits to be the requested
348 * group permission bits. This is not what we want, as it will
349 * mess up any inherited ACL bits that were set. JRA.
351 int saved_errno = errno; /* We may get ENOSYS */
352 if ((SMB_VFS_CHMOD_ACL(handle->conn, path, mode) == -1) && (errno == ENOSYS))
353 errno = saved_errno;
356 return result;
359 static int cephwrap_rmdir(struct vfs_handle_struct *handle, const char *path)
361 int result;
363 DEBUG(10, ("[CEPH] rmdir(%p, %s)\n", handle, path));
364 result = ceph_rmdir(handle->data, path);
365 DEBUG(10, ("[CEPH] rmdir(...) = %d\n", result));
366 WRAP_RETURN(result);
369 static int cephwrap_closedir(struct vfs_handle_struct *handle, DIR *dirp)
371 int result;
373 DEBUG(10, ("[CEPH] closedir(%p, %p)\n", handle, dirp));
374 result = ceph_closedir(handle->data, (struct ceph_dir_result *) dirp);
375 DEBUG(10, ("[CEPH] closedir(...) = %d\n", result));
376 WRAP_RETURN(result);
379 /* File operations */
381 static int cephwrap_open(struct vfs_handle_struct *handle,
382 struct smb_filename *smb_fname,
383 files_struct *fsp, int flags, mode_t mode)
385 int result = -ENOENT;
386 DEBUG(10, ("[CEPH] open(%p, %s, %p, %d, %d)\n", handle, smb_fname_str_dbg(smb_fname), fsp, flags, mode));
388 if (smb_fname->stream_name) {
389 goto out;
392 result = ceph_open(handle->data, smb_fname->base_name, flags, mode);
393 out:
394 DEBUG(10, ("[CEPH] open(...) = %d\n", result));
395 WRAP_RETURN(result);
398 static int cephwrap_close(struct vfs_handle_struct *handle, files_struct *fsp)
400 int result;
402 DEBUG(10, ("[CEPH] close(%p, %p)\n", handle, fsp));
403 result = ceph_close(handle->data, fsp->fh->fd);
404 DEBUG(10, ("[CEPH] close(...) = %d\n", result));
406 WRAP_RETURN(result);
409 static ssize_t cephwrap_read(struct vfs_handle_struct *handle, files_struct *fsp, void *data, size_t n)
411 ssize_t result;
413 DEBUG(10, ("[CEPH] read(%p, %p, %p, %llu)\n", handle, fsp, data, llu(n)));
415 /* Using -1 for the offset means read/write rather than pread/pwrite */
416 result = ceph_read(handle->data, fsp->fh->fd, data, n, -1);
417 DEBUG(10, ("[CEPH] read(...) = %llu\n", llu(result)));
418 WRAP_RETURN(result);
421 static ssize_t cephwrap_pread(struct vfs_handle_struct *handle, files_struct *fsp, void *data,
422 size_t n, off_t offset)
424 ssize_t result;
426 DEBUG(10, ("[CEPH] pread(%p, %p, %p, %llu, %llu)\n", handle, fsp, data, llu(n), llu(offset)));
428 result = ceph_read(handle->data, fsp->fh->fd, data, n, offset);
429 DEBUG(10, ("[CEPH] pread(...) = %llu\n", llu(result)));
430 WRAP_RETURN(result);
434 static ssize_t cephwrap_write(struct vfs_handle_struct *handle, files_struct *fsp, const void *data, size_t n)
436 ssize_t result;
438 DEBUG(10, ("[CEPH] write(%p, %p, %p, %llu)\n", handle, fsp, data, llu(n)));
440 result = ceph_write(handle->data, fsp->fh->fd, data, n, -1);
442 DEBUG(10, ("[CEPH] write(...) = %llu\n", llu(result)));
443 if (result < 0) {
444 WRAP_RETURN(result);
446 fsp->fh->pos += result;
447 return result;
450 static ssize_t cephwrap_pwrite(struct vfs_handle_struct *handle, files_struct *fsp, const void *data,
451 size_t n, off_t offset)
453 ssize_t result;
455 DEBUG(10, ("[CEPH] pwrite(%p, %p, %p, %llu, %llu)\n", handle, fsp, data, llu(n), llu(offset)));
456 result = ceph_write(handle->data, fsp->fh->fd, data, n, offset);
457 DEBUG(10, ("[CEPH] pwrite(...) = %llu\n", llu(result)));
458 WRAP_RETURN(result);
461 static off_t cephwrap_lseek(struct vfs_handle_struct *handle, files_struct *fsp, off_t offset, int whence)
463 off_t result = 0;
465 DEBUG(10, ("[CEPH] cephwrap_lseek\n"));
466 /* Cope with 'stat' file opens. */
467 if (fsp->fh->fd != -1) {
468 result = ceph_lseek(handle->data, fsp->fh->fd, offset, whence);
470 WRAP_RETURN(result);
473 static ssize_t cephwrap_sendfile(struct vfs_handle_struct *handle, int tofd, files_struct *fromfsp, const DATA_BLOB *hdr,
474 off_t offset, size_t n)
477 * We cannot support sendfile because libceph is in user space.
479 DEBUG(10, ("[CEPH] cephwrap_sendfile\n"));
480 errno = ENOTSUP;
481 return -1;
484 static ssize_t cephwrap_recvfile(struct vfs_handle_struct *handle,
485 int fromfd,
486 files_struct *tofsp,
487 off_t offset,
488 size_t n)
491 * We cannot support recvfile because libceph is in user space.
493 DEBUG(10, ("[CEPH] cephwrap_recvfile\n"));
494 errno=ENOTSUP;
495 return -1;
498 static int cephwrap_rename(struct vfs_handle_struct *handle,
499 const struct smb_filename *smb_fname_src,
500 const struct smb_filename *smb_fname_dst)
502 int result = -1;
503 DEBUG(10, ("[CEPH] cephwrap_rename\n"));
504 if (smb_fname_src->stream_name || smb_fname_dst->stream_name) {
505 errno = ENOENT;
506 return result;
509 result = ceph_rename(handle->data, smb_fname_src->base_name, smb_fname_dst->base_name);
510 WRAP_RETURN(result);
513 static int cephwrap_fsync(struct vfs_handle_struct *handle, files_struct *fsp)
515 int result;
516 DEBUG(10, ("[CEPH] cephwrap_fsync\n"));
517 result = ceph_fsync(handle->data, fsp->fh->fd, false);
518 WRAP_RETURN(result);
521 static void cephwrap_init_stat_ex_from_stat(struct stat_ex *dst, const struct stat *src)
523 ZERO_STRUCT(*dst);
525 dst->st_ex_dev = src->st_dev;
526 dst->st_ex_ino = src->st_ino;
527 dst->st_ex_mode = src->st_mode;
528 dst->st_ex_nlink = src->st_nlink;
529 dst->st_ex_uid = src->st_uid;
530 dst->st_ex_gid = src->st_gid;
531 dst->st_ex_rdev = src->st_rdev;
532 dst->st_ex_size = src->st_size;
533 dst->st_ex_atime.tv_sec = src->st_atime;
534 dst->st_ex_mtime.tv_sec = src->st_mtime;
535 dst->st_ex_ctime.tv_sec = src->st_ctime;
536 dst->st_ex_btime.tv_sec = src->st_mtime;
537 dst->st_ex_blksize = src->st_blksize;
538 dst->st_ex_blocks = src->st_blocks;
541 static int cephwrap_stat(struct vfs_handle_struct *handle,
542 struct smb_filename *smb_fname)
544 int result = -1;
545 struct stat stbuf;
547 DEBUG(10, ("[CEPH] stat(%p, %s)\n", handle, smb_fname_str_dbg(smb_fname)));
549 if (smb_fname->stream_name) {
550 errno = ENOENT;
551 return result;
554 result = ceph_stat(handle->data, smb_fname->base_name, (struct stat *) &stbuf);
555 DEBUG(10, ("[CEPH] stat(...) = %d\n", result));
556 if (result < 0) {
557 WRAP_RETURN(result);
558 } else {
559 DEBUG(10, ("[CEPH]\tstbuf = {dev = %llu, ino = %llu, mode = 0x%x, nlink = %llu, "
560 "uid = %d, gid = %d, rdev = %llu, size = %llu, blksize = %llu, "
561 "blocks = %llu, atime = %llu, mtime = %llu, ctime = %llu}\n",
562 llu(stbuf.st_dev), llu(stbuf.st_ino), stbuf.st_mode, llu(stbuf.st_nlink),
563 stbuf.st_uid, stbuf.st_gid, llu(stbuf.st_rdev), llu(stbuf.st_size), llu(stbuf.st_blksize),
564 llu(stbuf.st_blocks), llu(stbuf.st_atime), llu(stbuf.st_mtime), llu(stbuf.st_ctime)));
566 cephwrap_init_stat_ex_from_stat(&(smb_fname->st), &stbuf);
567 DEBUG(10, ("[CEPH] mode = 0x%x\n", smb_fname->st.st_ex_mode));
568 return result;
571 static int cephwrap_fstat(struct vfs_handle_struct *handle, files_struct *fsp, SMB_STRUCT_STAT *sbuf)
573 int result = -1;
574 struct stat stbuf;
576 DEBUG(10, ("[CEPH] fstat(%p, %d)\n", handle, fsp->fh->fd));
577 result = ceph_fstat(handle->data, fsp->fh->fd, (struct stat *) &stbuf);
578 DEBUG(10, ("[CEPH] fstat(...) = %d\n", result));
579 if (result < 0) {
580 WRAP_RETURN(result);
581 } else {
582 DEBUG(10, ("[CEPH]\tstbuf = {dev = %llu, ino = %llu, mode = 0x%x, nlink = %llu, "
583 "uid = %d, gid = %d, rdev = %llu, size = %llu, blksize = %llu, "
584 "blocks = %llu, atime = %llu, mtime = %llu, ctime = %llu}\n",
585 llu(stbuf.st_dev), llu(stbuf.st_ino), stbuf.st_mode, llu(stbuf.st_nlink),
586 stbuf.st_uid, stbuf.st_gid, llu(stbuf.st_rdev), llu(stbuf.st_size), llu(stbuf.st_blksize),
587 llu(stbuf.st_blocks), llu(stbuf.st_atime), llu(stbuf.st_mtime), llu(stbuf.st_ctime)));
590 cephwrap_init_stat_ex_from_stat(sbuf, &stbuf);
591 DEBUG(10, ("[CEPH] mode = 0x%x\n", sbuf->st_ex_mode));
592 return result;
595 static int cephwrap_lstat(struct vfs_handle_struct *handle,
596 struct smb_filename *smb_fname)
598 int result = -1;
599 struct stat stbuf;
601 DEBUG(10, ("[CEPH] lstat(%p, %s)\n", handle, smb_fname_str_dbg(smb_fname)));
603 if (smb_fname->stream_name) {
604 errno = ENOENT;
605 return result;
608 result = ceph_lstat(handle->data, smb_fname->base_name, &stbuf);
609 DEBUG(10, ("[CEPH] lstat(...) = %d\n", result));
610 if (result < 0) {
611 WRAP_RETURN(result);
613 cephwrap_init_stat_ex_from_stat(&(smb_fname->st), &stbuf);
614 return result;
617 static int cephwrap_unlink(struct vfs_handle_struct *handle,
618 const struct smb_filename *smb_fname)
620 int result = -1;
622 DEBUG(10, ("[CEPH] unlink(%p, %s)\n", handle, smb_fname_str_dbg(smb_fname)));
623 if (smb_fname->stream_name) {
624 errno = ENOENT;
625 return result;
627 result = ceph_unlink(handle->data, smb_fname->base_name);
628 DEBUG(10, ("[CEPH] unlink(...) = %d\n", result));
629 WRAP_RETURN(result);
632 static int cephwrap_chmod(struct vfs_handle_struct *handle, const char *path, mode_t mode)
634 int result;
636 DEBUG(10, ("[CEPH] chmod(%p, %s, %d)\n", handle, path, mode));
639 * We need to do this due to the fact that the default POSIX ACL
640 * chmod modifies the ACL *mask* for the group owner, not the
641 * group owner bits directly. JRA.
646 int saved_errno = errno; /* We might get ENOSYS */
647 if ((result = SMB_VFS_CHMOD_ACL(handle->conn, path, mode)) == 0) {
648 return result;
650 /* Error - return the old errno. */
651 errno = saved_errno;
654 result = ceph_chmod(handle->data, path, mode);
655 DEBUG(10, ("[CEPH] chmod(...) = %d\n", result));
656 WRAP_RETURN(result);
659 static int cephwrap_fchmod(struct vfs_handle_struct *handle, files_struct *fsp, mode_t mode)
661 int result;
663 DEBUG(10, ("[CEPH] fchmod(%p, %p, %d)\n", handle, fsp, mode));
666 * We need to do this due to the fact that the default POSIX ACL
667 * chmod modifies the ACL *mask* for the group owner, not the
668 * group owner bits directly. JRA.
672 int saved_errno = errno; /* We might get ENOSYS */
673 if ((result = SMB_VFS_FCHMOD_ACL(fsp, mode)) == 0) {
674 return result;
676 /* Error - return the old errno. */
677 errno = saved_errno;
680 #if defined(HAVE_FCHMOD)
681 result = ceph_fchmod(handle->data, fsp->fh->fd, mode);
682 DEBUG(10, ("[CEPH] fchmod(...) = %d\n", result));
683 WRAP_RETURN(result);
684 #else
685 errno = ENOSYS;
686 #endif
687 return -1;
690 static int cephwrap_chown(struct vfs_handle_struct *handle, const char *path, uid_t uid, gid_t gid)
692 int result;
693 DEBUG(10, ("[CEPH] chown(%p, %s, %d, %d)\n", handle, path, uid, gid));
694 result = ceph_chown(handle->data, path, uid, gid);
695 DEBUG(10, ("[CEPH] chown(...) = %d\n", result));
696 WRAP_RETURN(result);
699 static int cephwrap_fchown(struct vfs_handle_struct *handle, files_struct *fsp, uid_t uid, gid_t gid)
701 int result;
702 #ifdef HAVE_FCHOWN
704 DEBUG(10, ("[CEPH] fchown(%p, %p, %d, %d)\n", handle, fsp, uid, gid));
705 result = ceph_fchown(handle->data, fsp->fh->fd, uid, gid);
706 DEBUG(10, ("[CEPH] fchown(...) = %d\n", result));
707 WRAP_RETURN(result);
708 #else
709 errno = ENOSYS;
710 result = -1;
711 #endif
712 return result;
715 static int cephwrap_lchown(struct vfs_handle_struct *handle, const char *path, uid_t uid, gid_t gid)
717 int result;
719 DEBUG(10, ("[CEPH] lchown(%p, %s, %d, %d)\n", handle, path, uid, gid));
720 result = ceph_lchown(handle->data, path, uid, gid);
721 DEBUG(10, ("[CEPH] lchown(...) = %d\n", result));
722 WRAP_RETURN(result);
725 static int cephwrap_chdir(struct vfs_handle_struct *handle, const char *path)
727 int result = -1;
728 DEBUG(10, ("[CEPH] chdir(%p, %s)\n", handle, path));
730 * If the path is just / use chdir because Ceph is below / and
731 * cannot deal with changing directory above its mount point
733 if (path && !strcmp(path, "/"))
734 return chdir(path);
736 result = ceph_chdir(handle->data, path);
737 DEBUG(10, ("[CEPH] chdir(...) = %d\n", result));
738 WRAP_RETURN(result);
741 static char *cephwrap_getwd(struct vfs_handle_struct *handle)
743 const char *cwd = ceph_getcwd(handle->data);
744 DEBUG(10, ("[CEPH] getwd(%p) = %s\n", handle, cwd));
745 return SMB_STRDUP(cwd);
748 static int cephwrap_ntimes(struct vfs_handle_struct *handle,
749 const struct smb_filename *smb_fname,
750 struct smb_file_time *ft)
752 struct utimbuf buf;
753 buf.actime = ft->atime.tv_sec;
754 buf.modtime = ft->mtime.tv_sec;
755 int result = ceph_utime(handle->data, smb_fname->base_name, &buf);
756 DEBUG(10, ("[CEPH] ntimes(%p, %s, {%ld, %ld, %ld, %ld}) = %d\n", handle, smb_fname_str_dbg(smb_fname),
757 ft->mtime.tv_sec, ft->atime.tv_sec, ft->ctime.tv_sec,
758 ft->create_time.tv_sec, result));
759 return result;
762 static int strict_allocate_ftruncate(struct vfs_handle_struct *handle, files_struct *fsp, off_t len)
764 off_t space_to_write;
765 uint64_t space_avail;
766 uint64_t bsize,dfree,dsize;
767 int ret;
768 NTSTATUS status;
769 SMB_STRUCT_STAT *pst;
771 status = vfs_stat_fsp(fsp);
772 if (!NT_STATUS_IS_OK(status)) {
773 return -1;
775 pst = &fsp->fsp_name->st;
777 #ifdef S_ISFIFO
778 if (S_ISFIFO(pst->st_ex_mode))
779 return 0;
780 #endif
782 if (pst->st_ex_size == len)
783 return 0;
785 /* Shrink - just ftruncate. */
786 if (pst->st_ex_size > len)
787 return ftruncate(fsp->fh->fd, len);
789 space_to_write = len - pst->st_ex_size;
791 /* for allocation try fallocate first. This can fail on some
792 platforms e.g. when the filesystem doesn't support it and no
793 emulation is being done by the libc (like on AIX with JFS1). In that
794 case we do our own emulation. fallocate implementations can
795 return ENOTSUP or EINVAL in cases like that. */
796 ret = SMB_VFS_FALLOCATE(fsp, VFS_FALLOCATE_EXTEND_SIZE,
797 pst->st_ex_size, space_to_write);
798 if (ret == -1 && errno == ENOSPC) {
799 return -1;
801 if (ret == 0) {
802 return 0;
804 DEBUG(10,("[CEPH] strict_allocate_ftruncate: SMB_VFS_FALLOCATE failed with "
805 "error %d. Falling back to slow manual allocation\n", errno));
807 /* available disk space is enough or not? */
808 space_avail = get_dfree_info(fsp->conn,
809 fsp->fsp_name->base_name, false,
810 &bsize,&dfree,&dsize);
811 /* space_avail is 1k blocks */
812 if (space_avail == (uint64_t)-1 ||
813 ((uint64_t)space_to_write/1024 > space_avail) ) {
814 errno = ENOSPC;
815 return -1;
818 /* Write out the real space on disk. */
819 return vfs_slow_fallocate(fsp, pst->st_ex_size, space_to_write);
822 static int cephwrap_ftruncate(struct vfs_handle_struct *handle, files_struct *fsp, off_t len)
824 int result = -1;
825 SMB_STRUCT_STAT st;
826 char c = 0;
827 off_t currpos;
829 DEBUG(10, ("[CEPH] ftruncate(%p, %p, %llu\n", handle, fsp, llu(len)));
831 if (lp_strict_allocate(SNUM(fsp->conn))) {
832 result = strict_allocate_ftruncate(handle, fsp, len);
833 return result;
836 /* we used to just check HAVE_FTRUNCATE_EXTEND and only use
837 sys_ftruncate if the system supports it. Then I discovered that
838 you can have some filesystems that support ftruncate
839 expansion and some that don't! On Linux fat can't do
840 ftruncate extend but ext2 can. */
842 result = ceph_ftruncate(handle->data, fsp->fh->fd, len);
843 if (result == 0)
844 goto done;
846 /* According to W. R. Stevens advanced UNIX prog. Pure 4.3 BSD cannot
847 extend a file with ftruncate. Provide alternate implementation
848 for this */
849 currpos = SMB_VFS_LSEEK(fsp, 0, SEEK_CUR);
850 if (currpos == -1) {
851 goto done;
854 /* Do an fstat to see if the file is longer than the requested
855 size in which case the ftruncate above should have
856 succeeded or shorter, in which case seek to len - 1 and
857 write 1 byte of zero */
858 if (SMB_VFS_FSTAT(fsp, &st) == -1) {
859 goto done;
862 #ifdef S_ISFIFO
863 if (S_ISFIFO(st.st_ex_mode)) {
864 result = 0;
865 goto done;
867 #endif
869 if (st.st_ex_size == len) {
870 result = 0;
871 goto done;
874 if (st.st_ex_size > len) {
875 /* the sys_ftruncate should have worked */
876 goto done;
879 if (SMB_VFS_LSEEK(fsp, len-1, SEEK_SET) != len -1)
880 goto done;
882 if (SMB_VFS_WRITE(fsp, &c, 1)!=1)
883 goto done;
885 /* Seek to where we were */
886 if (SMB_VFS_LSEEK(fsp, currpos, SEEK_SET) != currpos)
887 goto done;
888 result = 0;
890 done:
892 return result;
895 static bool cephwrap_lock(struct vfs_handle_struct *handle, files_struct *fsp, int op, off_t offset, off_t count, int type)
897 DEBUG(10, ("[CEPH] lock\n"));
898 return true;
901 static int cephwrap_kernel_flock(struct vfs_handle_struct *handle, files_struct *fsp,
902 uint32 share_mode, uint32 access_mask)
904 DEBUG(10, ("[CEPH] kernel_flock\n"));
906 * We must return zero here and pretend all is good.
907 * One day we might have this in CEPH.
909 return 0;
912 static bool cephwrap_getlock(struct vfs_handle_struct *handle, files_struct *fsp, off_t *poffset, off_t *pcount, int *ptype, pid_t *ppid)
914 DEBUG(10, ("[CEPH] getlock returning false and errno=0\n"));
916 errno = 0;
917 return false;
921 * We cannot let this fall through to the default, because the file might only
922 * be accessible from libceph (which is a user-space client) but the fd might
923 * be for some file the kernel knows about.
925 static int cephwrap_linux_setlease(struct vfs_handle_struct *handle, files_struct *fsp,
926 int leasetype)
928 int result = -1;
930 DEBUG(10, ("[CEPH] linux_setlease\n"));
931 errno = ENOSYS;
932 return result;
935 static int cephwrap_symlink(struct vfs_handle_struct *handle, const char *oldpath, const char *newpath)
937 int result = -1;
938 DEBUG(10, ("[CEPH] symlink(%p, %s, %s)\n", handle, oldpath, newpath));
939 result = ceph_symlink(handle->data, oldpath, newpath);
940 DEBUG(10, ("[CEPH] symlink(...) = %d\n", result));
941 WRAP_RETURN(result);
944 static int cephwrap_readlink(struct vfs_handle_struct *handle, const char *path, char *buf, size_t bufsiz)
946 int result = -1;
947 DEBUG(10, ("[CEPH] readlink(%p, %s, %p, %llu)\n", handle, path, buf, llu(bufsiz)));
948 result = ceph_readlink(handle->data, path, buf, bufsiz);
949 DEBUG(10, ("[CEPH] readlink(...) = %d\n", result));
950 WRAP_RETURN(result);
953 static int cephwrap_link(struct vfs_handle_struct *handle, const char *oldpath, const char *newpath)
955 int result = -1;
956 DEBUG(10, ("[CEPH] link(%p, %s, %s)\n", handle, oldpath, newpath));
957 result = ceph_link(handle->data, oldpath, newpath);
958 DEBUG(10, ("[CEPH] link(...) = %d\n", result));
959 WRAP_RETURN(result);
962 static int cephwrap_mknod(struct vfs_handle_struct *handle, const char *pathname, mode_t mode, SMB_DEV_T dev)
964 int result = -1;
965 DEBUG(10, ("[CEPH] mknod(%p, %s)\n", handle, pathname));
966 result = ceph_mknod(handle->data, pathname, mode, dev);
967 DEBUG(10, ("[CEPH] mknod(...) = %d\n", result));
968 WRAP_RETURN(result);
972 * This is a simple version of real-path ... a better version is needed to
973 * ask libceph about symbolic links.
975 static char *cephwrap_realpath(struct vfs_handle_struct *handle, const char *path)
977 char *result;
978 size_t len = strlen(path);
980 result = SMB_MALLOC_ARRAY(char, PATH_MAX+1);
981 if (len && (path[0] == '/')) {
982 int r = asprintf(&result, "%s", path);
983 if (r < 0) return NULL;
984 } else if ((len >= 2) && (path[0] == '.') && (path[1] == '/')) {
985 if (len == 2) {
986 int r = asprintf(&result, "%s",
987 handle->conn->connectpath);
988 if (r < 0) return NULL;
989 } else {
990 int r = asprintf(&result, "%s/%s",
991 handle->conn->connectpath, &path[2]);
992 if (r < 0) return NULL;
994 } else {
995 int r = asprintf(&result, "%s/%s",
996 handle->conn->connectpath, path);
997 if (r < 0) return NULL;
999 DEBUG(10, ("[CEPH] realpath(%p, %s) = %s\n", handle, path, result));
1000 return result;
1003 static NTSTATUS cephwrap_notify_watch(struct vfs_handle_struct *vfs_handle,
1004 struct sys_notify_context *ctx,
1005 const char *path,
1006 uint32_t *filter,
1007 uint32_t *subdir_filter,
1008 void (*callback)(struct sys_notify_context *ctx,
1009 void *private_data,
1010 struct notify_event *ev),
1011 void *private_data,
1012 void *handle_p)
1015 * We cannot call inotify on files the kernel does not know about
1017 return NT_STATUS_OK;
1020 static int cephwrap_chflags(struct vfs_handle_struct *handle, const char *path,
1021 unsigned int flags)
1023 errno = ENOSYS;
1024 return -1;
1027 static int cephwrap_get_real_filename(struct vfs_handle_struct *handle,
1028 const char *path,
1029 const char *name,
1030 TALLOC_CTX *mem_ctx,
1031 char **found_name)
1034 * Don't fall back to get_real_filename so callers can differentiate
1035 * between a full directory scan and an actual case-insensitive stat.
1037 errno = EOPNOTSUPP;
1038 return -1;
1041 static const char *cephwrap_connectpath(struct vfs_handle_struct *handle,
1042 const char *fname)
1044 return handle->conn->connectpath;
1047 /****************************************************************
1048 Extended attribute operations.
1049 *****************************************************************/
1051 static ssize_t cephwrap_getxattr(struct vfs_handle_struct *handle,const char *path, const char *name, void *value, size_t size)
1053 DEBUG(10, ("[CEPH] getxattr(%p, %s, %s, %p, %llu)\n", handle, path, name, value, llu(size)));
1054 int ret = ceph_getxattr(handle->data, path, name, value, size);
1055 DEBUG(10, ("[CEPH] getxattr(...) = %d\n", ret));
1056 if (ret < 0) {
1057 WRAP_RETURN(ret);
1058 } else {
1059 return (ssize_t)ret;
1063 static ssize_t cephwrap_fgetxattr(struct vfs_handle_struct *handle, struct files_struct *fsp, const char *name, void *value, size_t size)
1065 DEBUG(10, ("[CEPH] fgetxattr(%p, %p, %s, %p, %llu)\n", handle, fsp, name, value, llu(size)));
1066 int ret = ceph_getxattr(handle->data, fsp->fsp_name->base_name, name, value, size);
1067 DEBUG(10, ("[CEPH] fgetxattr(...) = %d\n", ret));
1068 if (ret < 0) {
1069 WRAP_RETURN(ret);
1070 } else {
1071 return (ssize_t)ret;
1075 static ssize_t cephwrap_listxattr(struct vfs_handle_struct *handle, const char *path, char *list, size_t size)
1077 DEBUG(10, ("[CEPH] listxattr(%p, %s, %p, %llu)\n", handle, path, list, llu(size)));
1078 int ret = ceph_listxattr(handle->data, path, list, size);
1079 DEBUG(10, ("[CEPH] listxattr(...) = %d\n", ret));
1080 if (ret < 0) {
1081 WRAP_RETURN(ret);
1082 } else {
1083 return (ssize_t)ret;
1087 #if 0
1088 static ssize_t cephwrap_llistxattr(struct vfs_handle_struct *handle, const char *path, char *list, size_t size)
1090 DEBUG(10, ("[CEPH] llistxattr(%p, %s, %p, %llu)\n", handle, path, list, llu(size)));
1091 int ret = ceph_llistxattr(handle->data, path, list, size);
1092 DEBUG(10, ("[CEPH] listxattr(...) = %d\n", ret));
1093 if (ret < 0) {
1094 WRAP_RETURN(ret);
1095 } else {
1096 return (ssize_t)ret;
1099 #endif
1101 static ssize_t cephwrap_flistxattr(struct vfs_handle_struct *handle, struct files_struct *fsp, char *list, size_t size)
1103 DEBUG(10, ("[CEPH] flistxattr(%p, %p, %s, %llu)\n", handle, fsp, list, llu(size)));
1104 int ret = ceph_listxattr(handle->data, fsp->fsp_name->base_name, list, size);
1105 DEBUG(10, ("[CEPH] flistxattr(...) = %d\n", ret));
1106 if (ret < 0) {
1107 WRAP_RETURN(ret);
1108 } else {
1109 return (ssize_t)ret;
1113 static int cephwrap_removexattr(struct vfs_handle_struct *handle, const char *path, const char *name)
1115 DEBUG(10, ("[CEPH] removexattr(%p, %s, %s)\n", handle, path, name));
1116 int ret = ceph_removexattr(handle->data, path, name);
1117 DEBUG(10, ("[CEPH] removexattr(...) = %d\n", ret));
1118 WRAP_RETURN(ret);
1121 static int cephwrap_fremovexattr(struct vfs_handle_struct *handle, struct files_struct *fsp, const char *name)
1123 DEBUG(10, ("[CEPH] fremovexattr(%p, %p, %s)\n", handle, fsp, name));
1124 int ret = ceph_removexattr(handle->data, fsp->fsp_name->base_name, name);
1125 DEBUG(10, ("[CEPH] fremovexattr(...) = %d\n", ret));
1126 WRAP_RETURN(ret);
1129 static int cephwrap_setxattr(struct vfs_handle_struct *handle, const char *path, const char *name, const void *value, size_t size, int flags)
1131 DEBUG(10, ("[CEPH] setxattr(%p, %s, %s, %p, %llu, %d)\n", handle, path, name, value, llu(size), flags));
1132 int ret = ceph_setxattr(handle->data, path, name, value, size, flags);
1133 DEBUG(10, ("[CEPH] setxattr(...) = %d\n", ret));
1134 WRAP_RETURN(ret);
1137 static int cephwrap_fsetxattr(struct vfs_handle_struct *handle, struct files_struct *fsp, const char *name, const void *value, size_t size, int flags)
1139 DEBUG(10, ("[CEPH] fsetxattr(%p, %p, %s, %p, %llu, %d)\n", handle, fsp, name, value, llu(size), flags));
1140 int ret = ceph_setxattr(handle->data, fsp->fsp_name->base_name, name, value, size, flags);
1141 DEBUG(10, ("[CEPH] fsetxattr(...) = %d\n", ret));
1142 WRAP_RETURN(ret);
1145 static bool cephwrap_aio_force(struct vfs_handle_struct *handle, struct files_struct *fsp)
1149 * We do not support AIO yet.
1152 DEBUG(10, ("[CEPH] cephwrap_aio_force(%p, %p) = false (errno = ENOTSUP)\n", handle, fsp));
1153 errno = ENOTSUP;
1154 return false;
1157 static bool cephwrap_is_offline(struct vfs_handle_struct *handle,
1158 const struct smb_filename *fname,
1159 SMB_STRUCT_STAT *sbuf)
1161 return false;
1164 static int cephwrap_set_offline(struct vfs_handle_struct *handle,
1165 const struct smb_filename *fname)
1167 errno = ENOTSUP;
1168 return -1;
1171 static struct vfs_fn_pointers ceph_fns = {
1172 /* Disk operations */
1174 .connect_fn = cephwrap_connect,
1175 .disconnect_fn = cephwrap_disconnect,
1176 .disk_free_fn = cephwrap_disk_free,
1177 .get_quota_fn = cephwrap_get_quota,
1178 .set_quota_fn = cephwrap_set_quota,
1179 .statvfs_fn = cephwrap_statvfs,
1181 /* Directory operations */
1183 .opendir_fn = cephwrap_opendir,
1184 .fdopendir_fn = cephwrap_fdopendir,
1185 .readdir_fn = cephwrap_readdir,
1186 .seekdir_fn = cephwrap_seekdir,
1187 .telldir_fn = cephwrap_telldir,
1188 .rewind_dir_fn = cephwrap_rewinddir,
1189 .mkdir_fn = cephwrap_mkdir,
1190 .rmdir_fn = cephwrap_rmdir,
1191 .closedir_fn = cephwrap_closedir,
1193 /* File operations */
1195 .open_fn = cephwrap_open,
1196 .close_fn = cephwrap_close,
1197 .read_fn = cephwrap_read,
1198 .pread_fn = cephwrap_pread,
1199 .write_fn = cephwrap_write,
1200 .pwrite_fn = cephwrap_pwrite,
1201 .lseek_fn = cephwrap_lseek,
1202 .sendfile_fn = cephwrap_sendfile,
1203 .recvfile_fn = cephwrap_recvfile,
1204 .rename_fn = cephwrap_rename,
1205 .fsync_fn = cephwrap_fsync,
1206 .stat_fn = cephwrap_stat,
1207 .fstat_fn = cephwrap_fstat,
1208 .lstat_fn = cephwrap_lstat,
1209 .unlink_fn = cephwrap_unlink,
1210 .chmod_fn = cephwrap_chmod,
1211 .fchmod_fn = cephwrap_fchmod,
1212 .chown_fn = cephwrap_chown,
1213 .fchown_fn = cephwrap_fchown,
1214 .lchown_fn = cephwrap_lchown,
1215 .chdir_fn = cephwrap_chdir,
1216 .getwd_fn = cephwrap_getwd,
1217 .ntimes_fn = cephwrap_ntimes,
1218 .ftruncate_fn = cephwrap_ftruncate,
1219 .lock_fn = cephwrap_lock,
1220 .kernel_flock_fn = cephwrap_kernel_flock,
1221 .linux_setlease_fn = cephwrap_linux_setlease,
1222 .getlock_fn = cephwrap_getlock,
1223 .symlink_fn = cephwrap_symlink,
1224 .readlink_fn = cephwrap_readlink,
1225 .link_fn = cephwrap_link,
1226 .mknod_fn = cephwrap_mknod,
1227 .realpath_fn = cephwrap_realpath,
1228 .notify_watch_fn = cephwrap_notify_watch,
1229 .chflags_fn = cephwrap_chflags,
1230 .get_real_filename_fn = cephwrap_get_real_filename,
1231 .connectpath_fn = cephwrap_connectpath,
1233 /* EA operations. */
1234 .getxattr_fn = cephwrap_getxattr,
1235 .fgetxattr_fn = cephwrap_fgetxattr,
1236 .listxattr_fn = cephwrap_listxattr,
1237 .flistxattr_fn = cephwrap_flistxattr,
1238 .removexattr_fn = cephwrap_removexattr,
1239 .fremovexattr_fn = cephwrap_fremovexattr,
1240 .setxattr_fn = cephwrap_setxattr,
1241 .fsetxattr_fn = cephwrap_fsetxattr,
1243 /* aio operations */
1244 .aio_force_fn = cephwrap_aio_force,
1246 /* offline operations */
1247 .is_offline_fn = cephwrap_is_offline,
1248 .set_offline_fn = cephwrap_set_offline
1251 NTSTATUS vfs_ceph_init(void);
1252 NTSTATUS vfs_ceph_init(void)
1254 return smb_register_vfs(SMB_VFS_INTERFACE_VERSION,
1255 "ceph", &ceph_fns);