source3/modules/vfs_ceph.c

   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>
   8
   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.
  13
  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.
  18
  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/>.
  21 */
  22
  23 /*
  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.
  26  *
  27  * Add the following smb.conf parameter to each share that will be hosted on
  28  * Ceph:
  29  *
  30  *   vfs objects = ceph [any others you need go here]
  31  */
  32
  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"
  42
  43 #undef DBGC_CLASS
  44 #define DBGC_CLASS DBGC_VFS
  45
  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
  50
  51 /*
  52  * Use %llu whenever we have a 64bit unsigned int, and cast to (long long unsigned)
  53  */
  54 #define llu(_var) ((long long unsigned)_var)
  55
  56 /*
  57  * Note, libceph's return code model is to return -errno! So we have to convert
  58  * to what Samba expects, with is set errno to -return and return -1
  59  */
  60 #define WRAP_RETURN(_res) \
  61         errno = 0; \
  62         if (_res < 0) { \
  63                 errno = -_res; \
  64                 return -1; \
  65         } \
  66         return _res \
  67
  68 /*
  69  * We mount only one file system and then all shares are assumed to be in that.
  70  * FIXME: If we want to support more than one FS, then we have to deal with
  71  * this differently.
  72  *
  73  * So, cmount tells us if we have been this way before and whether
  74  * we need to mount ceph and cmount_cnt tells us how many times we have
  75  * connected
  76  */
  77 static struct ceph_mount_info * cmount = NULL;
  78 static uint32_t cmount_cnt = 0;
  79
  80 /* Check for NULL pointer parameters in cephwrap_* functions */
  81
  82 /* We don't want to have NULL function pointers lying around.  Someone
  83    is sure to try and execute them.  These stubs are used to prevent
  84    this possibility. */
  85
  86 static int cephwrap_connect(struct vfs_handle_struct *handle,  const char *service, const char *user)
  87 {
  88         int ret;
  89         char buf[256];
  90         int snum = SNUM(handle->conn);
  91         const char *conf_file;
  92         const char *user_id;
  93
  94         if (cmount) {
  95                 handle->data = cmount; /* We have been here before */
  96                 cmount_cnt++;
  97                 return 0;
  98         }
  99
 100         /* if config_file and/or user_id are NULL, ceph will use defaults */
 101         conf_file = lp_parm_const_string(snum, "ceph", "config_file", NULL);
 102         user_id = lp_parm_const_string(snum, "ceph", "user_id", NULL);
 103
 104         DBG_DEBUG("[CEPH] calling: ceph_create\n");
 105         ret = ceph_create(&cmount, user_id);
 106         if (ret) {
 107                 goto err_out;
 108         }
 109
 110         DBG_DEBUG("[CEPH] calling: ceph_conf_read_file with %s\n",
 111                   (conf_file == NULL ? "default path" : conf_file));
 112         ret = ceph_conf_read_file(cmount, conf_file);
 113         if (ret) {
 114                 goto err_cm_release;
 115         }
 116
 117         DBG_DEBUG("[CEPH] calling: ceph_conf_get\n");
 118         ret = ceph_conf_get(cmount, "log file", buf, sizeof(buf));
 119         if (ret < 0) {
 120                 goto err_cm_release;
 121         }
 122
 123         /* libcephfs disables POSIX ACL support by default, enable it... */
 124         ret = ceph_conf_set(cmount, "client_acl_type", "posix_acl");
 125         if (ret < 0) {
 126                 goto err_cm_release;
 127         }
 128         /* tell libcephfs to perform local permission checks */
 129         ret = ceph_conf_set(cmount, "fuse_default_permissions", "false");
 130         if (ret < 0) {
 131                 goto err_cm_release;
 132         }
 133
 134         DBG_DEBUG("[CEPH] calling: ceph_mount\n");
 135         ret = ceph_mount(cmount, NULL);
 136         if (ret < 0) {
 137                 goto err_cm_release;
 138         }
 139
 140         /*
 141          * encode mount context/state into our vfs/connection holding structure
 142          * cmount is a ceph_mount_t*
 143          */
 144         handle->data = cmount;
 145         cmount_cnt++;
 146
 147         /*
 148          * Unless we have an async implementation of getxattrat turn this off.
 149          */
 150         lp_do_parameter(SNUM(handle->conn), "smbd async dosmode", "false");
 151
 152         return 0;
 153
 154 err_cm_release:
 155         ceph_release(cmount);
 156         cmount = NULL;
 157 err_out:
 158         /*
 159          * Handle the error correctly. Ceph returns -errno.
 160          */
 161         DBG_DEBUG("[CEPH] Error return: %s\n", strerror(-ret));
 162         WRAP_RETURN(ret);
 163 }
 164
 165 static void cephwrap_disconnect(struct vfs_handle_struct *handle)
 166 {
 167         int ret;
 168
 169         if (!cmount) {
 170                 DBG_ERR("[CEPH] Error, ceph not mounted\n");
 171                 return;
 172         }
 173
 174         /* Should we unmount/shutdown? Only if the last disconnect? */
 175         if (--cmount_cnt) {
 176                 DBG_DEBUG("[CEPH] Not shuting down CEPH because still more connections\n");
 177                 return;
 178         }
 179
 180         ret = ceph_unmount(cmount);
 181         if (ret < 0) {
 182                 DBG_ERR("[CEPH] failed to unmount: %s\n", strerror(-ret));
 183         }
 184
 185         ret = ceph_release(cmount);
 186         if (ret < 0) {
 187                 DBG_ERR("[CEPH] failed to release: %s\n", strerror(-ret));
 188         }
 189
 190         cmount = NULL;  /* Make it safe */
 191 }
 192
 193 /* Disk operations */
 194
 195 static uint64_t cephwrap_disk_free(struct vfs_handle_struct *handle,
 196                                 const struct smb_filename *smb_fname,
 197                                 uint64_t *bsize,
 198                                 uint64_t *dfree,
 199                                 uint64_t *dsize)
 200 {
 201         struct statvfs statvfs_buf;
 202         int ret;
 203
 204         if (!(ret = ceph_statfs(handle->data, smb_fname->base_name,
 205                         &statvfs_buf))) {
 206                 /*
 207                  * Provide all the correct values.
 208                  */
 209                 *bsize = statvfs_buf.f_bsize;
 210                 *dfree = statvfs_buf.f_bavail;
 211                 *dsize = statvfs_buf.f_blocks;
 212                 DBG_DEBUG("[CEPH] bsize: %llu, dfree: %llu, dsize: %llu\n",
 213                         llu(*bsize), llu(*dfree), llu(*dsize));
 214                 return *dfree;
 215         } else {
 216                 DBG_DEBUG("[CEPH] ceph_statfs returned %d\n", ret);
 217                 WRAP_RETURN(ret);
 218         }
 219 }
 220
 221 static int cephwrap_get_quota(struct vfs_handle_struct *handle,
 222                                 const struct smb_filename *smb_fname,
 223                                 enum SMB_QUOTA_TYPE qtype,
 224                                 unid_t id,
 225                                 SMB_DISK_QUOTA *qt)
 226 {
 227         /* libceph: Ceph does not implement this */
 228 #if 0
 229 /* was ifdef HAVE_SYS_QUOTAS */
 230         int ret;
 231
 232         ret = ceph_get_quota(handle->conn->connectpath, qtype, id, qt);
 233
 234         if (ret) {
 235                 errno = -ret;
 236                 ret = -1;
 237         }
 238
 239         return ret;
 240 #else
 241         errno = ENOSYS;
 242         return -1;
 243 #endif
 244 }
 245
 246 static int cephwrap_set_quota(struct vfs_handle_struct *handle,  enum SMB_QUOTA_TYPE qtype, unid_t id, SMB_DISK_QUOTA *qt)
 247 {
 248         /* libceph: Ceph does not implement this */
 249 #if 0
 250 /* was ifdef HAVE_SYS_QUOTAS */
 251         int ret;
 252
 253         ret = ceph_set_quota(handle->conn->connectpath, qtype, id, qt);
 254         if (ret) {
 255                 errno = -ret;
 256                 ret = -1;
 257         }
 258
 259         return ret;
 260 #else
 261         WRAP_RETURN(-ENOSYS);
 262 #endif
 263 }
 264
 265 static int cephwrap_statvfs(struct vfs_handle_struct *handle,
 266                                 const struct smb_filename *smb_fname,
 267                                 vfs_statvfs_struct *statbuf)
 268 {
 269         struct statvfs statvfs_buf;
 270         int ret;
 271
 272         ret = ceph_statfs(handle->data, smb_fname->base_name, &statvfs_buf);
 273         if (ret < 0) {
 274                 WRAP_RETURN(ret);
 275         }
 276
 277         statbuf->OptimalTransferSize = statvfs_buf.f_frsize;
 278         statbuf->BlockSize = statvfs_buf.f_bsize;
 279         statbuf->TotalBlocks = statvfs_buf.f_blocks;
 280         statbuf->BlocksAvail = statvfs_buf.f_bfree;
 281         statbuf->UserBlocksAvail = statvfs_buf.f_bavail;
 282         statbuf->TotalFileNodes = statvfs_buf.f_files;
 283         statbuf->FreeFileNodes = statvfs_buf.f_ffree;
 284         statbuf->FsIdentifier = statvfs_buf.f_fsid;
 285         DBG_DEBUG("[CEPH] f_bsize: %ld, f_blocks: %ld, f_bfree: %ld, f_bavail: %ld\n",
 286                 (long int)statvfs_buf.f_bsize, (long int)statvfs_buf.f_blocks,
 287                 (long int)statvfs_buf.f_bfree, (long int)statvfs_buf.f_bavail);
 288
 289         return ret;
 290 }
 291
 292 static uint32_t cephwrap_fs_capabilities(struct vfs_handle_struct *handle,
 293                                          enum timestamp_set_resolution *p_ts_res)
 294 {
 295         uint32_t caps = FILE_CASE_SENSITIVE_SEARCH | FILE_CASE_PRESERVED_NAMES;
 296
 297         *p_ts_res = TIMESTAMP_SET_NT_OR_BETTER;
 298
 299         return caps;
 300 }
 301
 302 /* Directory operations */
 303
 304 static DIR *cephwrap_opendir(struct vfs_handle_struct *handle,
 305                              const struct smb_filename *smb_fname,
 306                              const char *mask, uint32_t attr)
 307 {
 308         int ret = 0;
 309         struct ceph_dir_result *result;
 310         DBG_DEBUG("[CEPH] opendir(%p, %s)\n", handle, smb_fname->base_name);
 311
 312         /* Returns NULL if it does not exist or there are problems ? */
 313         ret = ceph_opendir(handle->data, smb_fname->base_name, &result);
 314         if (ret < 0) {
 315                 result = NULL;
 316                 errno = -ret; /* We return result which is NULL in this case */
 317         }
 318
 319         DBG_DEBUG("[CEPH] opendir(...) = %d\n", ret);
 320         return (DIR *) result;
 321 }
 322
 323 static DIR *cephwrap_fdopendir(struct vfs_handle_struct *handle,
 324                                struct files_struct *fsp,
 325                                const char *mask,
 326                                uint32_t attributes)
 327 {
 328         /* OpenDir_fsp() falls back to regular open */
 329         errno = ENOSYS;
 330         return NULL;
 331 }
 332
 333 static struct dirent *cephwrap_readdir(struct vfs_handle_struct *handle,
 334                                        DIR *dirp,
 335                                        SMB_STRUCT_STAT *sbuf)
 336 {
 337         struct dirent *result;
 338
 339         DBG_DEBUG("[CEPH] readdir(%p, %p)\n", handle, dirp);
 340         result = ceph_readdir(handle->data, (struct ceph_dir_result *) dirp);
 341         DBG_DEBUG("[CEPH] readdir(...) = %p\n", result);
 342
 343         /* Default Posix readdir() does not give us stat info.
 344          * Set to invalid to indicate we didn't return this info. */
 345         if (sbuf)
 346                 SET_STAT_INVALID(*sbuf);
 347         return result;
 348 }
 349
 350 static void cephwrap_seekdir(struct vfs_handle_struct *handle, DIR *dirp, long offset)
 351 {
 352         DBG_DEBUG("[CEPH] seekdir(%p, %p, %ld)\n", handle, dirp, offset);
 353         ceph_seekdir(handle->data, (struct ceph_dir_result *) dirp, offset);
 354 }
 355
 356 static long cephwrap_telldir(struct vfs_handle_struct *handle, DIR *dirp)
 357 {
 358         long ret;
 359         DBG_DEBUG("[CEPH] telldir(%p, %p)\n", handle, dirp);
 360         ret = ceph_telldir(handle->data, (struct ceph_dir_result *) dirp);
 361         DBG_DEBUG("[CEPH] telldir(...) = %ld\n", ret);
 362         WRAP_RETURN(ret);
 363 }
 364
 365 static void cephwrap_rewinddir(struct vfs_handle_struct *handle, DIR *dirp)
 366 {
 367         DBG_DEBUG("[CEPH] rewinddir(%p, %p)\n", handle, dirp);
 368         ceph_rewinddir(handle->data, (struct ceph_dir_result *) dirp);
 369 }
 370
 371 static int cephwrap_mkdirat(struct vfs_handle_struct *handle,
 372                         files_struct *dirfsp,
 373                         const struct smb_filename *smb_fname,
 374                         mode_t mode)
 375 {
 376         int result;
 377         char *parent = NULL;
 378         const char *path = smb_fname->base_name;
 379
 380         DBG_DEBUG("[CEPH] mkdir(%p, %s)\n", handle, path);
 381
 382         SMB_ASSERT(dirfsp == dirfsp->conn->cwd_fsp);
 383
 384         if (lp_inherit_acls(SNUM(handle->conn))
 385             && parent_dirname(talloc_tos(), path, &parent, NULL)
 386             && directory_has_default_acl(handle->conn, parent)) {
 387                 mode = 0777;
 388         }
 389
 390         TALLOC_FREE(parent);
 391
 392         result = ceph_mkdir(handle->data, path, mode);
 393         return WRAP_RETURN(result);
 394 }
 395
 396 static int cephwrap_closedir(struct vfs_handle_struct *handle, DIR *dirp)
 397 {
 398         int result;
 399
 400         DBG_DEBUG("[CEPH] closedir(%p, %p)\n", handle, dirp);
 401         result = ceph_closedir(handle->data, (struct ceph_dir_result *) dirp);
 402         DBG_DEBUG("[CEPH] closedir(...) = %d\n", result);
 403         WRAP_RETURN(result);
 404 }
 405
 406 /* File operations */
 407
 408 static int cephwrap_open(struct vfs_handle_struct *handle,
 409                         struct smb_filename *smb_fname,
 410                         files_struct *fsp, int flags, mode_t mode)
 411 {
 412         int result = -ENOENT;
 413         DBG_DEBUG("[CEPH] open(%p, %s, %p, %d, %d)\n", handle,
 414                   smb_fname_str_dbg(smb_fname), fsp, flags, mode);
 415
 416         if (smb_fname->stream_name) {
 417                 goto out;
 418         }
 419
 420         result = ceph_open(handle->data, smb_fname->base_name, flags, mode);
 421 out:
 422         DBG_DEBUG("[CEPH] open(...) = %d\n", result);
 423         WRAP_RETURN(result);
 424 }
 425
 426 static int cephwrap_close(struct vfs_handle_struct *handle, files_struct *fsp)
 427 {
 428         int result;
 429
 430         DBG_DEBUG("[CEPH] close(%p, %p)\n", handle, fsp);
 431         result = ceph_close(handle->data, fsp->fh->fd);
 432         DBG_DEBUG("[CEPH] close(...) = %d\n", result);
 433
 434         WRAP_RETURN(result);
 435 }
 436
 437 static ssize_t cephwrap_pread(struct vfs_handle_struct *handle, files_struct *fsp, void *data,
 438                         size_t n, off_t offset)
 439 {
 440         ssize_t result;
 441
 442         DBG_DEBUG("[CEPH] pread(%p, %p, %p, %llu, %llu)\n", handle, fsp, data, llu(n), llu(offset));
 443
 444         result = ceph_read(handle->data, fsp->fh->fd, data, n, offset);
 445         DBG_DEBUG("[CEPH] pread(...) = %llu\n", llu(result));
 446         WRAP_RETURN(result);
 447 }
 448
 449 struct cephwrap_pread_state {
 450         ssize_t bytes_read;
 451         struct vfs_aio_state vfs_aio_state;
 452 };
 453
 454 /*
 455  * Fake up an async ceph read by calling the synchronous API.
 456  */
 457 static struct tevent_req *cephwrap_pread_send(struct vfs_handle_struct *handle,
 458                                               TALLOC_CTX *mem_ctx,
 459                                               struct tevent_context *ev,
 460                                               struct files_struct *fsp,
 461                                               void *data,
 462                                               size_t n, off_t offset)
 463 {
 464         struct tevent_req *req = NULL;
 465         struct cephwrap_pread_state *state = NULL;
 466         int ret = -1;
 467
 468         DBG_DEBUG("[CEPH] %s\n", __func__);
 469         req = tevent_req_create(mem_ctx, &state, struct cephwrap_pread_state);
 470         if (req == NULL) {
 471                 return NULL;
 472         }
 473
 474         ret = ceph_read(handle->data, fsp->fh->fd, data, n, offset);
 475         if (ret < 0) {
 476                 /* ceph returns -errno on error. */
 477                 tevent_req_error(req, -ret);
 478                 return tevent_req_post(req, ev);
 479         }
 480
 481         state->bytes_read = ret;
 482         tevent_req_done(req);
 483         /* Return and schedule the completion of the call. */
 484         return tevent_req_post(req, ev);
 485 }
 486
 487 static ssize_t cephwrap_pread_recv(struct tevent_req *req,
 488                                    struct vfs_aio_state *vfs_aio_state)
 489 {
 490         struct cephwrap_pread_state *state =
 491                 tevent_req_data(req, struct cephwrap_pread_state);
 492
 493         DBG_DEBUG("[CEPH] %s\n", __func__);
 494         if (tevent_req_is_unix_error(req, &vfs_aio_state->error)) {
 495                 return -1;
 496         }
 497         *vfs_aio_state = state->vfs_aio_state;
 498         return state->bytes_read;
 499 }
 500
 501 static ssize_t cephwrap_pwrite(struct vfs_handle_struct *handle, files_struct *fsp, const void *data,
 502                         size_t n, off_t offset)
 503 {
 504         ssize_t result;
 505
 506         DBG_DEBUG("[CEPH] pwrite(%p, %p, %p, %llu, %llu)\n", handle, fsp, data, llu(n), llu(offset));
 507         result = ceph_write(handle->data, fsp->fh->fd, data, n, offset);
 508         DBG_DEBUG("[CEPH] pwrite(...) = %llu\n", llu(result));
 509         WRAP_RETURN(result);
 510 }
 511
 512 struct cephwrap_pwrite_state {
 513         ssize_t bytes_written;
 514         struct vfs_aio_state vfs_aio_state;
 515 };
 516
 517 /*
 518  * Fake up an async ceph write by calling the synchronous API.
 519  */
 520 static struct tevent_req *cephwrap_pwrite_send(struct vfs_handle_struct *handle,
 521                                                TALLOC_CTX *mem_ctx,
 522                                                struct tevent_context *ev,
 523                                                struct files_struct *fsp,
 524                                                const void *data,
 525                                                size_t n, off_t offset)
 526 {
 527         struct tevent_req *req = NULL;
 528         struct cephwrap_pwrite_state *state = NULL;
 529         int ret = -1;
 530
 531         DBG_DEBUG("[CEPH] %s\n", __func__);
 532         req = tevent_req_create(mem_ctx, &state, struct cephwrap_pwrite_state);
 533         if (req == NULL) {
 534                 return NULL;
 535         }
 536
 537         ret = ceph_write(handle->data, fsp->fh->fd, data, n, offset);
 538         if (ret < 0) {
 539                 /* ceph returns -errno on error. */
 540                 tevent_req_error(req, -ret);
 541                 return tevent_req_post(req, ev);
 542         }
 543
 544         state->bytes_written = ret;
 545         tevent_req_done(req);
 546         /* Return and schedule the completion of the call. */
 547         return tevent_req_post(req, ev);
 548 }
 549
 550 static ssize_t cephwrap_pwrite_recv(struct tevent_req *req,
 551                                     struct vfs_aio_state *vfs_aio_state)
 552 {
 553         struct cephwrap_pwrite_state *state =
 554                 tevent_req_data(req, struct cephwrap_pwrite_state);
 555
 556         DBG_DEBUG("[CEPH] %s\n", __func__);
 557         if (tevent_req_is_unix_error(req, &vfs_aio_state->error)) {
 558                 return -1;
 559         }
 560         *vfs_aio_state = state->vfs_aio_state;
 561         return state->bytes_written;
 562 }
 563
 564 static off_t cephwrap_lseek(struct vfs_handle_struct *handle, files_struct *fsp, off_t offset, int whence)
 565 {
 566         off_t result = 0;
 567
 568         DBG_DEBUG("[CEPH] cephwrap_lseek\n");
 569         result = ceph_lseek(handle->data, fsp->fh->fd, offset, whence);
 570         WRAP_RETURN(result);
 571 }
 572
 573 static ssize_t cephwrap_sendfile(struct vfs_handle_struct *handle, int tofd, files_struct *fromfsp, const DATA_BLOB *hdr,
 574                         off_t offset, size_t n)
 575 {
 576         /*
 577          * We cannot support sendfile because libceph is in user space.
 578          */
 579         DBG_DEBUG("[CEPH] cephwrap_sendfile\n");
 580         errno = ENOTSUP;
 581         return -1;
 582 }
 583
 584 static ssize_t cephwrap_recvfile(struct vfs_handle_struct *handle,
 585                         int fromfd,
 586                         files_struct *tofsp,
 587                         off_t offset,
 588                         size_t n)
 589 {
 590         /*
 591          * We cannot support recvfile because libceph is in user space.
 592          */
 593         DBG_DEBUG("[CEPH] cephwrap_recvfile\n");
 594         errno=ENOTSUP;
 595         return -1;
 596 }
 597
 598 static int cephwrap_renameat(struct vfs_handle_struct *handle,
 599                         files_struct *srcfsp,
 600                         const struct smb_filename *smb_fname_src,
 601                         files_struct *dstfsp,
 602                         const struct smb_filename *smb_fname_dst)
 603 {
 604         int result = -1;
 605         DBG_DEBUG("[CEPH] cephwrap_renameat\n");
 606         if (smb_fname_src->stream_name || smb_fname_dst->stream_name) {
 607                 errno = ENOENT;
 608                 return result;
 609         }
 610
 611         SMB_ASSERT(srcfsp == srcfsp->conn->cwd_fsp);
 612         SMB_ASSERT(dstfsp == dstfsp->conn->cwd_fsp);
 613
 614         result = ceph_rename(handle->data, smb_fname_src->base_name, smb_fname_dst->base_name);
 615         WRAP_RETURN(result);
 616 }
 617
 618 /*
 619  * Fake up an async ceph fsync by calling the synchronous API.
 620  */
 621
 622 static struct tevent_req *cephwrap_fsync_send(struct vfs_handle_struct *handle,
 623                                         TALLOC_CTX *mem_ctx,
 624                                         struct tevent_context *ev,
 625                                         files_struct *fsp)
 626 {
 627         struct tevent_req *req = NULL;
 628         struct vfs_aio_state *state = NULL;
 629         int ret = -1;
 630
 631         DBG_DEBUG("[CEPH] cephwrap_fsync_send\n");
 632
 633         req = tevent_req_create(mem_ctx, &state, struct vfs_aio_state);
 634         if (req == NULL) {
 635                 return NULL;
 636         }
 637
 638         /* Make sync call. */
 639         ret = ceph_fsync(handle->data, fsp->fh->fd, false);
 640
 641         if (ret != 0) {
 642                 /* ceph_fsync returns -errno on error. */
 643                 tevent_req_error(req, -ret);
 644                 return tevent_req_post(req, ev);
 645         }
 646
 647         /* Mark it as done. */
 648         tevent_req_done(req);
 649         /* Return and schedule the completion of the call. */
 650         return tevent_req_post(req, ev);
 651 }
 652
 653 static int cephwrap_fsync_recv(struct tevent_req *req,
 654                                 struct vfs_aio_state *vfs_aio_state)
 655 {
 656         struct vfs_aio_state *state =
 657                 tevent_req_data(req, struct vfs_aio_state);
 658
 659         DBG_DEBUG("[CEPH] cephwrap_fsync_recv\n");
 660
 661         if (tevent_req_is_unix_error(req, &vfs_aio_state->error)) {
 662                 return -1;
 663         }
 664         *vfs_aio_state = *state;
 665         return 0;
 666 }
 667
 668 #define SAMBA_STATX_ATTR_MASK   (CEPH_STATX_BASIC_STATS|CEPH_STATX_BTIME)
 669
 670 static void init_stat_ex_from_ceph_statx(struct stat_ex *dst, const struct ceph_statx *stx)
 671 {
 672         DBG_DEBUG("[CEPH]\tstx = {dev = %llx, ino = %llu, mode = 0x%x, "
 673                   "nlink = %llu, uid = %d, gid = %d, rdev = %llx, size = %llu, "
 674                   "blksize = %llu, blocks = %llu, atime = %llu, mtime = %llu, "
 675                   "ctime = %llu, btime = %llu}\n",
 676                   llu(stx->stx_dev), llu(stx->stx_ino), stx->stx_mode,
 677                   llu(stx->stx_nlink), stx->stx_uid, stx->stx_gid,
 678                   llu(stx->stx_rdev), llu(stx->stx_size), llu(stx->stx_blksize),
 679                   llu(stx->stx_blocks), llu(stx->stx_atime.tv_sec),
 680                   llu(stx->stx_mtime.tv_sec), llu(stx->stx_ctime.tv_sec),
 681                   llu(stx->stx_btime.tv_sec));
 682
 683         if ((stx->stx_mask & SAMBA_STATX_ATTR_MASK) != SAMBA_STATX_ATTR_MASK) {
 684                 DBG_WARNING("%s: stx->stx_mask is incorrect (wanted %x, got %x)",
 685                                 __func__, SAMBA_STATX_ATTR_MASK, stx->stx_mask);
 686         }
 687
 688         dst->st_ex_dev = stx->stx_dev;
 689         dst->st_ex_rdev = stx->stx_rdev;
 690         dst->st_ex_ino = stx->stx_ino;
 691         dst->st_ex_mode = stx->stx_mode;
 692         dst->st_ex_uid = stx->stx_uid;
 693         dst->st_ex_gid = stx->stx_gid;
 694         dst->st_ex_size = stx->stx_size;
 695         dst->st_ex_nlink = stx->stx_nlink;
 696         dst->st_ex_atime = stx->stx_atime;
 697         dst->st_ex_btime = stx->stx_btime;
 698         dst->st_ex_ctime = stx->stx_ctime;
 699         dst->st_ex_mtime = stx->stx_mtime;
 700         dst->st_ex_itime = dst->st_ex_btime;
 701         dst->st_ex_iflags = ST_EX_IFLAG_CALCULATED_ITIME;
 702         dst->st_ex_blksize = stx->stx_blksize;
 703         dst->st_ex_blocks = stx->stx_blocks;
 704         dst->st_ex_file_id = dst->st_ex_ino;
 705         dst->st_ex_iflags |= ST_EX_IFLAG_CALCULATED_FILE_ID;
 706 }
 707
 708 static int cephwrap_stat(struct vfs_handle_struct *handle,
 709                         struct smb_filename *smb_fname)
 710 {
 711         int result = -1;
 712         struct ceph_statx stx;
 713
 714         DBG_DEBUG("[CEPH] stat(%p, %s)\n", handle, smb_fname_str_dbg(smb_fname));
 715
 716         if (smb_fname->stream_name) {
 717                 errno = ENOENT;
 718                 return result;
 719         }
 720
 721         result = ceph_statx(handle->data, smb_fname->base_name, &stx,
 722                                 SAMBA_STATX_ATTR_MASK, 0);
 723         DBG_DEBUG("[CEPH] statx(...) = %d\n", result);
 724         if (result < 0) {
 725                 WRAP_RETURN(result);
 726         }
 727
 728         init_stat_ex_from_ceph_statx(&smb_fname->st, &stx);
 729         DBG_DEBUG("[CEPH] mode = 0x%x\n", smb_fname->st.st_ex_mode);
 730         return result;
 731 }
 732
 733 static int cephwrap_fstat(struct vfs_handle_struct *handle, files_struct *fsp, SMB_STRUCT_STAT *sbuf)
 734 {
 735         int result = -1;
 736         struct ceph_statx stx;
 737
 738         DBG_DEBUG("[CEPH] fstat(%p, %d)\n", handle, fsp->fh->fd);
 739         result = ceph_fstatx(handle->data, fsp->fh->fd, &stx,
 740                                 SAMBA_STATX_ATTR_MASK, 0);
 741         DBG_DEBUG("[CEPH] fstat(...) = %d\n", result);
 742         if (result < 0) {
 743                 WRAP_RETURN(result);
 744         }
 745
 746         init_stat_ex_from_ceph_statx(sbuf, &stx);
 747         DBG_DEBUG("[CEPH] mode = 0x%x\n", sbuf->st_ex_mode);
 748         return result;
 749 }
 750
 751 static int cephwrap_lstat(struct vfs_handle_struct *handle,
 752                          struct smb_filename *smb_fname)
 753 {
 754         int result = -1;
 755         struct ceph_statx stx;
 756
 757         DBG_DEBUG("[CEPH] lstat(%p, %s)\n", handle, smb_fname_str_dbg(smb_fname));
 758
 759         if (smb_fname->stream_name) {
 760                 errno = ENOENT;
 761                 return result;
 762         }
 763
 764         result = ceph_statx(handle->data, smb_fname->base_name, &stx,
 765                                 SAMBA_STATX_ATTR_MASK, AT_SYMLINK_NOFOLLOW);
 766         DBG_DEBUG("[CEPH] lstat(...) = %d\n", result);
 767         if (result < 0) {
 768                 WRAP_RETURN(result);
 769         }
 770
 771         init_stat_ex_from_ceph_statx(&smb_fname->st, &stx);
 772         return result;
 773 }
 774
 775 static int cephwrap_ntimes(struct vfs_handle_struct *handle,
 776                          const struct smb_filename *smb_fname,
 777                          struct smb_file_time *ft)
 778 {
 779         struct ceph_statx stx = { 0 };
 780         int result;
 781         int mask = 0;
 782
 783         if (!is_omit_timespec(&ft->atime)) {
 784                 stx.stx_atime = ft->atime;
 785                 mask |= CEPH_SETATTR_ATIME;
 786         }
 787         if (!is_omit_timespec(&ft->mtime)) {
 788                 stx.stx_mtime = ft->mtime;
 789                 mask |= CEPH_SETATTR_MTIME;
 790         }
 791         if (!is_omit_timespec(&ft->create_time)) {
 792                 stx.stx_btime = ft->create_time;
 793                 mask |= CEPH_SETATTR_BTIME;
 794         }
 795
 796         if (!mask) {
 797                 return 0;
 798         }
 799
 800         result = ceph_setattrx(handle->data, smb_fname->base_name, &stx, mask, 0);
 801         DBG_DEBUG("[CEPH] ntimes(%p, %s, {%ld, %ld, %ld, %ld}) = %d\n", handle, smb_fname_str_dbg(smb_fname),
 802                                 ft->mtime.tv_sec, ft->atime.tv_sec, ft->ctime.tv_sec,
 803                                 ft->create_time.tv_sec, result);
 804         return result;
 805 }
 806
 807 static int cephwrap_unlinkat(struct vfs_handle_struct *handle,
 808                         struct files_struct *dirfsp,
 809                         const struct smb_filename *smb_fname,
 810                         int flags)
 811 {
 812         int result = -1;
 813
 814         DBG_DEBUG("[CEPH] unlink(%p, %s)\n",
 815                 handle,
 816                 smb_fname_str_dbg(smb_fname));
 817         SMB_ASSERT(dirfsp == dirfsp->conn->cwd_fsp);
 818         if (smb_fname->stream_name) {
 819                 errno = ENOENT;
 820                 return result;
 821         }
 822         if (flags & AT_REMOVEDIR) {
 823                 result = ceph_rmdir(handle->data, smb_fname->base_name);
 824         } else {
 825                 result = ceph_unlink(handle->data, smb_fname->base_name);
 826         }
 827         DBG_DEBUG("[CEPH] unlink(...) = %d\n", result);
 828         WRAP_RETURN(result);
 829 }
 830
 831 static int cephwrap_chmod(struct vfs_handle_struct *handle,
 832                         const struct smb_filename *smb_fname,
 833                         mode_t mode)
 834 {
 835         int result;
 836
 837         DBG_DEBUG("[CEPH] chmod(%p, %s, %d)\n", handle, smb_fname->base_name, mode);
 838         result = ceph_chmod(handle->data, smb_fname->base_name, mode);
 839         DBG_DEBUG("[CEPH] chmod(...) = %d\n", result);
 840         WRAP_RETURN(result);
 841 }
 842
 843 static int cephwrap_fchmod(struct vfs_handle_struct *handle, files_struct *fsp, mode_t mode)
 844 {
 845         int result;
 846
 847         DBG_DEBUG("[CEPH] fchmod(%p, %p, %d)\n", handle, fsp, mode);
 848         result = ceph_fchmod(handle->data, fsp->fh->fd, mode);
 849         DBG_DEBUG("[CEPH] fchmod(...) = %d\n", result);
 850         WRAP_RETURN(result);
 851 }
 852
 853 static int cephwrap_fchown(struct vfs_handle_struct *handle, files_struct *fsp, uid_t uid, gid_t gid)
 854 {
 855         int result;
 856
 857         DBG_DEBUG("[CEPH] fchown(%p, %p, %d, %d)\n", handle, fsp, uid, gid);
 858         result = ceph_fchown(handle->data, fsp->fh->fd, uid, gid);
 859         DBG_DEBUG("[CEPH] fchown(...) = %d\n", result);
 860         WRAP_RETURN(result);
 861 }
 862
 863 static int cephwrap_lchown(struct vfs_handle_struct *handle,
 864                         const struct smb_filename *smb_fname,
 865                         uid_t uid,
 866                         gid_t gid)
 867 {
 868         int result;
 869         DBG_DEBUG("[CEPH] lchown(%p, %s, %d, %d)\n", handle, smb_fname->base_name, uid, gid);
 870         result = ceph_lchown(handle->data, smb_fname->base_name, uid, gid);
 871         DBG_DEBUG("[CEPH] lchown(...) = %d\n", result);
 872         WRAP_RETURN(result);
 873 }
 874
 875 static int cephwrap_chdir(struct vfs_handle_struct *handle,
 876                         const struct smb_filename *smb_fname)
 877 {
 878         int result = -1;
 879         DBG_DEBUG("[CEPH] chdir(%p, %s)\n", handle, smb_fname->base_name);
 880         result = ceph_chdir(handle->data, smb_fname->base_name);
 881         DBG_DEBUG("[CEPH] chdir(...) = %d\n", result);
 882         WRAP_RETURN(result);
 883 }
 884
 885 static struct smb_filename *cephwrap_getwd(struct vfs_handle_struct *handle,
 886                         TALLOC_CTX *ctx)
 887 {
 888         const char *cwd = ceph_getcwd(handle->data);
 889         DBG_DEBUG("[CEPH] getwd(%p) = %s\n", handle, cwd);
 890         return synthetic_smb_fname(ctx,
 891                                 cwd,
 892                                 NULL,
 893                                 NULL,
 894                                 0);
 895 }
 896
 897 static int strict_allocate_ftruncate(struct vfs_handle_struct *handle, files_struct *fsp, off_t len)
 898 {
 899         off_t space_to_write;
 900         int result;
 901         NTSTATUS status;
 902         SMB_STRUCT_STAT *pst;
 903
 904         status = vfs_stat_fsp(fsp);
 905         if (!NT_STATUS_IS_OK(status)) {
 906                 return -1;
 907         }
 908         pst = &fsp->fsp_name->st;
 909
 910 #ifdef S_ISFIFO
 911         if (S_ISFIFO(pst->st_ex_mode))
 912                 return 0;
 913 #endif
 914
 915         if (pst->st_ex_size == len)
 916                 return 0;
 917
 918         /* Shrink - just ftruncate. */
 919         if (pst->st_ex_size > len) {
 920                 result = ceph_ftruncate(handle->data, fsp->fh->fd, len);
 921                 WRAP_RETURN(result);
 922         }
 923
 924         space_to_write = len - pst->st_ex_size;
 925         result = ceph_fallocate(handle->data, fsp->fh->fd, 0, pst->st_ex_size,
 926                                 space_to_write);
 927         WRAP_RETURN(result);
 928 }
 929
 930 static int cephwrap_ftruncate(struct vfs_handle_struct *handle, files_struct *fsp, off_t len)
 931 {
 932         int result = -1;
 933
 934         DBG_DEBUG("[CEPH] ftruncate(%p, %p, %llu\n", handle, fsp, llu(len));
 935
 936         if (lp_strict_allocate(SNUM(fsp->conn))) {
 937                 return strict_allocate_ftruncate(handle, fsp, len);
 938         }
 939
 940         result = ceph_ftruncate(handle->data, fsp->fh->fd, len);
 941         WRAP_RETURN(result);
 942 }
 943
 944 static int cephwrap_fallocate(struct vfs_handle_struct *handle,
 945                               struct files_struct *fsp,
 946                               uint32_t mode,
 947                               off_t offset,
 948                               off_t len)
 949 {
 950         int result;
 951
 952         DBG_DEBUG("[CEPH] fallocate(%p, %p, %u, %llu, %llu\n",
 953                   handle, fsp, mode, llu(offset), llu(len));
 954         /* unsupported mode flags are rejected by libcephfs */
 955         result = ceph_fallocate(handle->data, fsp->fh->fd, mode, offset, len);
 956         DBG_DEBUG("[CEPH] fallocate(...) = %d\n", result);
 957         WRAP_RETURN(result);
 958 }
 959
 960 static bool cephwrap_lock(struct vfs_handle_struct *handle, files_struct *fsp, int op, off_t offset, off_t count, int type)
 961 {
 962         DBG_DEBUG("[CEPH] lock\n");
 963         return true;
 964 }
 965
 966 static int cephwrap_kernel_flock(struct vfs_handle_struct *handle,
 967                                  files_struct *fsp,
 968                                  uint32_t share_access,
 969                                  uint32_t access_mask)
 970 {
 971         DBG_ERR("[CEPH] flock unsupported! Consider setting "
 972                 "\"kernel share modes = no\"\n");
 973
 974         errno = ENOSYS;
 975         return -1;
 976 }
 977
 978 static int cephwrap_fcntl(vfs_handle_struct *handle,
 979                           files_struct *fsp, int cmd, va_list cmd_arg)
 980 {
 981         /*
 982          * SMB_VFS_FCNTL() is currently only called by vfs_set_blocking() to
 983          * clear O_NONBLOCK, etc for LOCK_MAND and FIFOs. Ignore it.
 984          */
 985         if (cmd == F_GETFL) {
 986                 return 0;
 987         } else if (cmd == F_SETFL) {
 988                 va_list dup_cmd_arg;
 989                 int opt;
 990
 991                 va_copy(dup_cmd_arg, cmd_arg);
 992                 opt = va_arg(dup_cmd_arg, int);
 993                 va_end(dup_cmd_arg);
 994                 if (opt == 0) {
 995                         return 0;
 996                 }
 997                 DBG_ERR("unexpected fcntl SETFL(%d)\n", opt);
 998                 goto err_out;
 999         }
1000         DBG_ERR("unexpected fcntl: %d\n", cmd);
1001 err_out:
1002         errno = EINVAL;
1003         return -1;
1004 }
1005
1006 static bool cephwrap_getlock(struct vfs_handle_struct *handle, files_struct *fsp, off_t *poffset, off_t *pcount, int *ptype, pid_t *ppid)
1007 {
1008         DBG_DEBUG("[CEPH] getlock returning false and errno=0\n");
1009
1010         errno = 0;
1011         return false;
1012 }
1013
1014 /*
1015  * We cannot let this fall through to the default, because the file might only
1016  * be accessible from libceph (which is a user-space client) but the fd might
1017  * be for some file the kernel knows about.
1018  */
1019 static int cephwrap_linux_setlease(struct vfs_handle_struct *handle, files_struct *fsp,
1020                                 int leasetype)
1021 {
1022         int result = -1;
1023
1024         DBG_DEBUG("[CEPH] linux_setlease\n");
1025         errno = ENOSYS;
1026         return result;
1027 }
1028
1029 static int cephwrap_symlinkat(struct vfs_handle_struct *handle,
1030                 const char *link_target,
1031                 struct files_struct *dirfsp,
1032                 const struct smb_filename *new_smb_fname)
1033 {
1034         int result = -1;
1035         DBG_DEBUG("[CEPH] symlink(%p, %s, %s)\n", handle,
1036                         link_target,
1037                         new_smb_fname->base_name);
1038
1039         SMB_ASSERT(dirfsp == dirfsp->conn->cwd_fsp);
1040
1041         result = ceph_symlink(handle->data,
1042                         link_target,
1043                         new_smb_fname->base_name);
1044         DBG_DEBUG("[CEPH] symlink(...) = %d\n", result);
1045         WRAP_RETURN(result);
1046 }
1047
1048 static int cephwrap_readlinkat(struct vfs_handle_struct *handle,
1049                 files_struct *dirfsp,
1050                 const struct smb_filename *smb_fname,
1051                 char *buf,
1052                 size_t bufsiz)
1053 {
1054         int result = -1;
1055         DBG_DEBUG("[CEPH] readlink(%p, %s, %p, %llu)\n", handle,
1056                         smb_fname->base_name, buf, llu(bufsiz));
1057
1058         SMB_ASSERT(dirfsp == dirfsp->conn->cwd_fsp);
1059
1060         result = ceph_readlink(handle->data, smb_fname->base_name, buf, bufsiz);
1061         DBG_DEBUG("[CEPH] readlink(...) = %d\n", result);
1062         WRAP_RETURN(result);
1063 }
1064
1065 static int cephwrap_linkat(struct vfs_handle_struct *handle,
1066                 files_struct *srcfsp,
1067                 const struct smb_filename *old_smb_fname,
1068                 files_struct *dstfsp,
1069                 const struct smb_filename *new_smb_fname,
1070                 int flags)
1071 {
1072         int result = -1;
1073         DBG_DEBUG("[CEPH] link(%p, %s, %s)\n", handle,
1074                         old_smb_fname->base_name,
1075                         new_smb_fname->base_name);
1076
1077         SMB_ASSERT(srcfsp == srcfsp->conn->cwd_fsp);
1078         SMB_ASSERT(dstfsp == dstfsp->conn->cwd_fsp);
1079
1080         result = ceph_link(handle->data,
1081                                 old_smb_fname->base_name,
1082                                 new_smb_fname->base_name);
1083         DBG_DEBUG("[CEPH] link(...) = %d\n", result);
1084         WRAP_RETURN(result);
1085 }
1086
1087 static int cephwrap_mknodat(struct vfs_handle_struct *handle,
1088                 files_struct *dirfsp,
1089                 const struct smb_filename *smb_fname,
1090                 mode_t mode,
1091                 SMB_DEV_T dev)
1092 {
1093         int result = -1;
1094         DBG_DEBUG("[CEPH] mknodat(%p, %s)\n", handle, smb_fname->base_name);
1095         SMB_ASSERT(dirfsp == dirfsp->conn->cwd_fsp);
1096         result = ceph_mknod(handle->data, smb_fname->base_name, mode, dev);
1097         DBG_DEBUG("[CEPH] mknodat(...) = %d\n", result);
1098         WRAP_RETURN(result);
1099 }
1100
1101 /*
1102  * This is a simple version of real-path ... a better version is needed to
1103  * ask libceph about symbolic links.
1104  */
1105 static struct smb_filename *cephwrap_realpath(struct vfs_handle_struct *handle,
1106                                 TALLOC_CTX *ctx,
1107                                 const struct smb_filename *smb_fname)
1108 {
1109         char *result = NULL;
1110         const char *path = smb_fname->base_name;
1111         size_t len = strlen(path);
1112         struct smb_filename *result_fname = NULL;
1113         int r = -1;
1114
1115         if (len && (path[0] == '/')) {
1116                 r = asprintf(&result, "%s", path);
1117         } else if ((len >= 2) && (path[0] == '.') && (path[1] == '/')) {
1118                 if (len == 2) {
1119                         r = asprintf(&result, "%s",
1120                                         handle->conn->cwd_fsp->fsp_name->base_name);
1121                 } else {
1122                         r = asprintf(&result, "%s/%s",
1123                                         handle->conn->cwd_fsp->fsp_name->base_name, &path[2]);
1124                 }
1125         } else {
1126                 r = asprintf(&result, "%s/%s",
1127                                 handle->conn->cwd_fsp->fsp_name->base_name, path);
1128         }
1129
1130         if (r < 0) {
1131                 return NULL;
1132         }
1133
1134         DBG_DEBUG("[CEPH] realpath(%p, %s) = %s\n", handle, path, result);
1135         result_fname = synthetic_smb_fname(ctx,
1136                                 result,
1137                                 NULL,
1138                                 NULL,
1139                                 0);
1140         SAFE_FREE(result);
1141         return result_fname;
1142 }
1143
1144 static int cephwrap_chflags(struct vfs_handle_struct *handle,
1145                         const struct smb_filename *smb_fname,
1146                         unsigned int flags)
1147 {
1148         errno = ENOSYS;
1149         return -1;
1150 }
1151
1152 static int cephwrap_get_real_filename(struct vfs_handle_struct *handle,
1153                                      const char *path,
1154                                      const char *name,
1155                                      TALLOC_CTX *mem_ctx,
1156                                      char **found_name)
1157 {
1158         /*
1159          * Don't fall back to get_real_filename so callers can differentiate
1160          * between a full directory scan and an actual case-insensitive stat.
1161          */
1162         errno = EOPNOTSUPP;
1163         return -1;
1164 }
1165
1166 static const char *cephwrap_connectpath(struct vfs_handle_struct *handle,
1167                                        const struct smb_filename *smb_fname)
1168 {
1169         return handle->conn->connectpath;
1170 }
1171
1172 /****************************************************************
1173  Extended attribute operations.
1174 *****************************************************************/
1175
1176 static ssize_t cephwrap_getxattr(struct vfs_handle_struct *handle,
1177                         const struct smb_filename *smb_fname,
1178                         const char *name,
1179                         void *value,
1180                         size_t size)
1181 {
1182         int ret;
1183         DBG_DEBUG("[CEPH] getxattr(%p, %s, %s, %p, %llu)\n", handle,
1184                         smb_fname->base_name, name, value, llu(size));
1185         ret = ceph_getxattr(handle->data,
1186                         smb_fname->base_name, name, value, size);
1187         DBG_DEBUG("[CEPH] getxattr(...) = %d\n", ret);
1188         if (ret < 0) {
1189                 WRAP_RETURN(ret);
1190         }
1191         return (ssize_t)ret;
1192 }
1193
1194 static ssize_t cephwrap_fgetxattr(struct vfs_handle_struct *handle, struct files_struct *fsp, const char *name, void *value, size_t size)
1195 {
1196         int ret;
1197         DBG_DEBUG("[CEPH] fgetxattr(%p, %p, %s, %p, %llu)\n", handle, fsp, name, value, llu(size));
1198         ret = ceph_fgetxattr(handle->data, fsp->fh->fd, name, value, size);
1199         DBG_DEBUG("[CEPH] fgetxattr(...) = %d\n", ret);
1200         if (ret < 0) {
1201                 WRAP_RETURN(ret);
1202         }
1203         return (ssize_t)ret;
1204 }
1205
1206 static ssize_t cephwrap_listxattr(struct vfs_handle_struct *handle,
1207                         const struct smb_filename *smb_fname,
1208                         char *list,
1209                         size_t size)
1210 {
1211         int ret;
1212         DBG_DEBUG("[CEPH] listxattr(%p, %s, %p, %llu)\n", handle,
1213                         smb_fname->base_name, list, llu(size));
1214         ret = ceph_listxattr(handle->data, smb_fname->base_name, list, size);
1215         DBG_DEBUG("[CEPH] listxattr(...) = %d\n", ret);
1216         if (ret < 0) {
1217                 WRAP_RETURN(ret);
1218         }
1219         return (ssize_t)ret;
1220 }
1221
1222 static ssize_t cephwrap_flistxattr(struct vfs_handle_struct *handle, struct files_struct *fsp, char *list, size_t size)
1223 {
1224         int ret;
1225         DBG_DEBUG("[CEPH] flistxattr(%p, %p, %p, %llu)\n",
1226                   handle, fsp, list, llu(size));
1227         ret = ceph_flistxattr(handle->data, fsp->fh->fd, list, size);
1228         DBG_DEBUG("[CEPH] flistxattr(...) = %d\n", ret);
1229         if (ret < 0) {
1230                 WRAP_RETURN(ret);
1231         }
1232         return (ssize_t)ret;
1233 }
1234
1235 static int cephwrap_removexattr(struct vfs_handle_struct *handle,
1236                                 const struct smb_filename *smb_fname,
1237                                 const char *name)
1238 {
1239         int ret;
1240         DBG_DEBUG("[CEPH] removexattr(%p, %s, %s)\n", handle,
1241                         smb_fname->base_name, name);
1242         ret = ceph_removexattr(handle->data, smb_fname->base_name, name);
1243         DBG_DEBUG("[CEPH] removexattr(...) = %d\n", ret);
1244         WRAP_RETURN(ret);
1245 }
1246
1247 static int cephwrap_fremovexattr(struct vfs_handle_struct *handle, struct files_struct *fsp, const char *name)
1248 {
1249         int ret;
1250         DBG_DEBUG("[CEPH] fremovexattr(%p, %p, %s)\n", handle, fsp, name);
1251         ret = ceph_fremovexattr(handle->data, fsp->fh->fd, name);
1252         DBG_DEBUG("[CEPH] fremovexattr(...) = %d\n", ret);
1253         WRAP_RETURN(ret);
1254 }
1255
1256 static int cephwrap_setxattr(struct vfs_handle_struct *handle,
1257                                 const struct smb_filename *smb_fname,
1258                                 const char *name,
1259                                 const void *value,
1260                                 size_t size,
1261                                 int flags)
1262 {
1263         int ret;
1264         DBG_DEBUG("[CEPH] setxattr(%p, %s, %s, %p, %llu, %d)\n", handle,
1265                         smb_fname->base_name, name, value, llu(size), flags);
1266         ret = ceph_setxattr(handle->data, smb_fname->base_name,
1267                         name, value, size, flags);
1268         DBG_DEBUG("[CEPH] setxattr(...) = %d\n", ret);
1269         WRAP_RETURN(ret);
1270 }
1271
1272 static int cephwrap_fsetxattr(struct vfs_handle_struct *handle, struct files_struct *fsp, const char *name, const void *value, size_t size, int flags)
1273 {
1274         int ret;
1275         DBG_DEBUG("[CEPH] fsetxattr(%p, %p, %s, %p, %llu, %d)\n", handle, fsp, name, value, llu(size), flags);
1276         ret = ceph_fsetxattr(handle->data, fsp->fh->fd,
1277                              name, value, size, flags);
1278         DBG_DEBUG("[CEPH] fsetxattr(...) = %d\n", ret);
1279         WRAP_RETURN(ret);
1280 }
1281
1282 static bool cephwrap_aio_force(struct vfs_handle_struct *handle, struct files_struct *fsp)
1283 {
1284
1285         /*
1286          * We do not support AIO yet.
1287          */
1288
1289         DBG_DEBUG("[CEPH] cephwrap_aio_force(%p, %p) = false (errno = ENOTSUP)\n", handle, fsp);
1290         errno = ENOTSUP;
1291         return false;
1292 }
1293
1294 static NTSTATUS cephwrap_create_dfs_pathat(struct vfs_handle_struct *handle,
1295                                 struct files_struct *dirfsp,
1296                                 const struct smb_filename *smb_fname,
1297                                 const struct referral *reflist,
1298                                 size_t referral_count)
1299 {
1300         TALLOC_CTX *frame = talloc_stackframe();
1301         NTSTATUS status = NT_STATUS_NO_MEMORY;
1302         int ret;
1303         char *msdfs_link = NULL;
1304
1305         SMB_ASSERT(dirfsp == dirfsp->conn->cwd_fsp);
1306
1307         /* Form the msdfs_link contents */
1308         msdfs_link = msdfs_link_string(frame,
1309                                         reflist,
1310                                         referral_count);
1311         if (msdfs_link == NULL) {
1312                 goto out;
1313         }
1314
1315         ret = ceph_symlink(handle->data,
1316                         msdfs_link,
1317                         smb_fname->base_name);
1318         if (ret == 0) {
1319                 status = NT_STATUS_OK;
1320         } else {
1321                 status = map_nt_error_from_unix(-ret);
1322         }
1323
1324   out:
1325
1326         DBG_DEBUG("[CEPH] create_dfs_pathat(%s) = %s\n",
1327                         smb_fname->base_name,
1328                         nt_errstr(status));
1329
1330         TALLOC_FREE(frame);
1331         return status;
1332 }
1333
1334 /*
1335  * Read and return the contents of a DFS redirect given a
1336  * pathname. A caller can pass in NULL for ppreflist and
1337  * preferral_count but still determine if this was a
1338  * DFS redirect point by getting NT_STATUS_OK back
1339  * without incurring the overhead of reading and parsing
1340  * the referral contents.
1341  */
1342
1343 static NTSTATUS cephwrap_read_dfs_pathat(struct vfs_handle_struct *handle,
1344                                 TALLOC_CTX *mem_ctx,
1345                                 struct files_struct *dirfsp,
1346                                 const struct smb_filename *smb_fname,
1347                                 struct referral **ppreflist,
1348                                 size_t *preferral_count)
1349 {
1350         NTSTATUS status = NT_STATUS_NO_MEMORY;
1351         size_t bufsize;
1352         char *link_target = NULL;
1353         int referral_len;
1354         bool ok;
1355 #if defined(HAVE_BROKEN_READLINK)
1356         char link_target_buf[PATH_MAX];
1357 #else
1358         char link_target_buf[7];
1359 #endif
1360
1361         SMB_ASSERT(dirfsp == dirfsp->conn->cwd_fsp);
1362
1363         if (ppreflist == NULL && preferral_count == NULL) {
1364                 /*
1365                  * We're only checking if this is a DFS
1366                  * redirect. We don't need to return data.
1367                  */
1368                 bufsize = sizeof(link_target_buf);
1369                 link_target = link_target_buf;
1370         } else {
1371                 bufsize = PATH_MAX;
1372                 link_target = talloc_array(mem_ctx, char, bufsize);
1373                 if (!link_target) {
1374                         goto err;
1375                 }
1376         }
1377
1378         referral_len = ceph_readlink(handle->data,
1379                                 smb_fname->base_name,
1380                                 link_target,
1381                                 bufsize - 1);
1382         if (referral_len < 0) {
1383                 /* ceph errors are -errno. */
1384                 if (-referral_len == EINVAL) {
1385                         DBG_INFO("%s is not a link.\n",
1386                                 smb_fname->base_name);
1387                         status = NT_STATUS_OBJECT_TYPE_MISMATCH;
1388                 } else {
1389                         status = map_nt_error_from_unix(-referral_len);
1390                         DBG_ERR("Error reading "
1391                                 "msdfs link %s: %s\n",
1392                                 smb_fname->base_name,
1393                         strerror(errno));
1394                 }
1395                 goto err;
1396         }
1397         link_target[referral_len] = '\0';
1398
1399         DBG_INFO("%s -> %s\n",
1400                         smb_fname->base_name,
1401                         link_target);
1402
1403         if (!strnequal(link_target, "msdfs:", 6)) {
1404                 status = NT_STATUS_OBJECT_TYPE_MISMATCH;
1405                 goto err;
1406         }
1407
1408         if (ppreflist == NULL && preferral_count == NULL) {
1409                 /* Early return for checking if this is a DFS link. */
1410                 return NT_STATUS_OK;
1411         }
1412
1413         ok = parse_msdfs_symlink(mem_ctx,
1414                         lp_msdfs_shuffle_referrals(SNUM(handle->conn)),
1415                         link_target,
1416                         ppreflist,
1417                         preferral_count);
1418
1419         if (ok) {
1420                 status = NT_STATUS_OK;
1421         } else {
1422                 status = NT_STATUS_NO_MEMORY;
1423         }
1424
1425   err:
1426
1427         if (link_target != link_target_buf) {
1428                 TALLOC_FREE(link_target);
1429         }
1430         return status;
1431 }
1432
1433 static struct vfs_fn_pointers ceph_fns = {
1434         /* Disk operations */
1435
1436         .connect_fn = cephwrap_connect,
1437         .disconnect_fn = cephwrap_disconnect,
1438         .disk_free_fn = cephwrap_disk_free,
1439         .get_quota_fn = cephwrap_get_quota,
1440         .set_quota_fn = cephwrap_set_quota,
1441         .statvfs_fn = cephwrap_statvfs,
1442         .fs_capabilities_fn = cephwrap_fs_capabilities,
1443
1444         /* Directory operations */
1445
1446         .opendir_fn = cephwrap_opendir,
1447         .fdopendir_fn = cephwrap_fdopendir,
1448         .readdir_fn = cephwrap_readdir,
1449         .seekdir_fn = cephwrap_seekdir,
1450         .telldir_fn = cephwrap_telldir,
1451         .rewind_dir_fn = cephwrap_rewinddir,
1452         .mkdirat_fn = cephwrap_mkdirat,
1453         .closedir_fn = cephwrap_closedir,
1454
1455         /* File operations */
1456
1457         .create_dfs_pathat_fn = cephwrap_create_dfs_pathat,
1458         .read_dfs_pathat_fn = cephwrap_read_dfs_pathat,
1459         .open_fn = cephwrap_open,
1460         .close_fn = cephwrap_close,
1461         .pread_fn = cephwrap_pread,
1462         .pread_send_fn = cephwrap_pread_send,
1463         .pread_recv_fn = cephwrap_pread_recv,
1464         .pwrite_fn = cephwrap_pwrite,
1465         .pwrite_send_fn = cephwrap_pwrite_send,
1466         .pwrite_recv_fn = cephwrap_pwrite_recv,
1467         .lseek_fn = cephwrap_lseek,
1468         .sendfile_fn = cephwrap_sendfile,
1469         .recvfile_fn = cephwrap_recvfile,
1470         .renameat_fn = cephwrap_renameat,
1471         .fsync_send_fn = cephwrap_fsync_send,
1472         .fsync_recv_fn = cephwrap_fsync_recv,
1473         .stat_fn = cephwrap_stat,
1474         .fstat_fn = cephwrap_fstat,
1475         .lstat_fn = cephwrap_lstat,
1476         .unlinkat_fn = cephwrap_unlinkat,
1477         .chmod_fn = cephwrap_chmod,
1478         .fchmod_fn = cephwrap_fchmod,
1479         .fchown_fn = cephwrap_fchown,
1480         .lchown_fn = cephwrap_lchown,
1481         .chdir_fn = cephwrap_chdir,
1482         .getwd_fn = cephwrap_getwd,
1483         .ntimes_fn = cephwrap_ntimes,
1484         .ftruncate_fn = cephwrap_ftruncate,
1485         .fallocate_fn = cephwrap_fallocate,
1486         .lock_fn = cephwrap_lock,
1487         .kernel_flock_fn = cephwrap_kernel_flock,
1488         .fcntl_fn = cephwrap_fcntl,
1489         .linux_setlease_fn = cephwrap_linux_setlease,
1490         .getlock_fn = cephwrap_getlock,
1491         .symlinkat_fn = cephwrap_symlinkat,
1492         .readlinkat_fn = cephwrap_readlinkat,
1493         .linkat_fn = cephwrap_linkat,
1494         .mknodat_fn = cephwrap_mknodat,
1495         .realpath_fn = cephwrap_realpath,
1496         .chflags_fn = cephwrap_chflags,
1497         .get_real_filename_fn = cephwrap_get_real_filename,
1498         .connectpath_fn = cephwrap_connectpath,
1499
1500         /* EA operations. */
1501         .getxattr_fn = cephwrap_getxattr,
1502         .getxattrat_send_fn = vfs_not_implemented_getxattrat_send,
1503         .getxattrat_recv_fn = vfs_not_implemented_getxattrat_recv,
1504         .fgetxattr_fn = cephwrap_fgetxattr,
1505         .listxattr_fn = cephwrap_listxattr,
1506         .flistxattr_fn = cephwrap_flistxattr,
1507         .removexattr_fn = cephwrap_removexattr,
1508         .fremovexattr_fn = cephwrap_fremovexattr,
1509         .setxattr_fn = cephwrap_setxattr,
1510         .fsetxattr_fn = cephwrap_fsetxattr,
1511
1512         /* Posix ACL Operations */
1513         .sys_acl_get_file_fn = posixacl_xattr_acl_get_file,
1514         .sys_acl_get_fd_fn = posixacl_xattr_acl_get_fd,
1515         .sys_acl_blob_get_file_fn = posix_sys_acl_blob_get_file,
1516         .sys_acl_blob_get_fd_fn = posix_sys_acl_blob_get_fd,
1517         .sys_acl_set_file_fn = posixacl_xattr_acl_set_file,
1518         .sys_acl_set_fd_fn = posixacl_xattr_acl_set_fd,
1519         .sys_acl_delete_def_file_fn = posixacl_xattr_acl_delete_def_file,
1520
1521         /* aio operations */
1522         .aio_force_fn = cephwrap_aio_force,
1523 };
1524
1525 static_decl_vfs;
1526 NTSTATUS vfs_ceph_init(TALLOC_CTX *ctx)
1527 {
1528         return smb_register_vfs(SMB_VFS_INTERFACE_VERSION,
1529                                 "ceph", &ceph_fns);
1530 }