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_fdopendir(struct vfs_handle_struct *handle,
 305                                struct files_struct *fsp,
 306                                const char *mask,
 307                                uint32_t attributes)
 308 {
 309         int ret = 0;
 310         struct ceph_dir_result *result;
 311         DBG_DEBUG("[CEPH] fdopendir(%p, %p)\n", handle, fsp);
 312
 313         ret = ceph_opendir(handle->data, fsp->fsp_name->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] fdopendir(...) = %d\n", ret);
 320         return (DIR *) result;
 321 }
 322
 323 static struct dirent *cephwrap_readdir(struct vfs_handle_struct *handle,
 324                                        DIR *dirp,
 325                                        SMB_STRUCT_STAT *sbuf)
 326 {
 327         struct dirent *result;
 328
 329         DBG_DEBUG("[CEPH] readdir(%p, %p)\n", handle, dirp);
 330         result = ceph_readdir(handle->data, (struct ceph_dir_result *) dirp);
 331         DBG_DEBUG("[CEPH] readdir(...) = %p\n", result);
 332
 333         /* Default Posix readdir() does not give us stat info.
 334          * Set to invalid to indicate we didn't return this info. */
 335         if (sbuf)
 336                 SET_STAT_INVALID(*sbuf);
 337         return result;
 338 }
 339
 340 static void cephwrap_seekdir(struct vfs_handle_struct *handle, DIR *dirp, long offset)
 341 {
 342         DBG_DEBUG("[CEPH] seekdir(%p, %p, %ld)\n", handle, dirp, offset);
 343         ceph_seekdir(handle->data, (struct ceph_dir_result *) dirp, offset);
 344 }
 345
 346 static long cephwrap_telldir(struct vfs_handle_struct *handle, DIR *dirp)
 347 {
 348         long ret;
 349         DBG_DEBUG("[CEPH] telldir(%p, %p)\n", handle, dirp);
 350         ret = ceph_telldir(handle->data, (struct ceph_dir_result *) dirp);
 351         DBG_DEBUG("[CEPH] telldir(...) = %ld\n", ret);
 352         WRAP_RETURN(ret);
 353 }
 354
 355 static void cephwrap_rewinddir(struct vfs_handle_struct *handle, DIR *dirp)
 356 {
 357         DBG_DEBUG("[CEPH] rewinddir(%p, %p)\n", handle, dirp);
 358         ceph_rewinddir(handle->data, (struct ceph_dir_result *) dirp);
 359 }
 360
 361 static int cephwrap_mkdirat(struct vfs_handle_struct *handle,
 362                         files_struct *dirfsp,
 363                         const struct smb_filename *smb_fname,
 364                         mode_t mode)
 365 {
 366         int result;
 367         struct smb_filename *parent = NULL;
 368         bool ok;
 369
 370         DBG_DEBUG("[CEPH] mkdir(%p, %s)\n",
 371                   handle, smb_fname_str_dbg(smb_fname));
 372
 373         SMB_ASSERT(dirfsp == dirfsp->conn->cwd_fsp);
 374
 375         if (lp_inherit_acls(SNUM(handle->conn))) {
 376                 ok = parent_smb_fname(talloc_tos(), smb_fname, &parent, NULL);
 377                 if (ok && directory_has_default_acl(handle->conn,
 378                                 dirfsp,
 379                                 parent))
 380                 {
 381                         mode = 0777;
 382                 }
 383         }
 384
 385         TALLOC_FREE(parent);
 386
 387         result = ceph_mkdir(handle->data, smb_fname->base_name, mode);
 388         return WRAP_RETURN(result);
 389 }
 390
 391 static int cephwrap_closedir(struct vfs_handle_struct *handle, DIR *dirp)
 392 {
 393         int result;
 394
 395         DBG_DEBUG("[CEPH] closedir(%p, %p)\n", handle, dirp);
 396         result = ceph_closedir(handle->data, (struct ceph_dir_result *) dirp);
 397         DBG_DEBUG("[CEPH] closedir(...) = %d\n", result);
 398         WRAP_RETURN(result);
 399 }
 400
 401 /* File operations */
 402
 403 static int cephwrap_openat(struct vfs_handle_struct *handle,
 404                            const struct files_struct *dirfsp,
 405                            const struct smb_filename *smb_fname,
 406                            files_struct *fsp,
 407                            int flags,
 408                            mode_t mode)
 409 {
 410         int result = -ENOENT;
 411
 412         /*
 413          * cephfs API doesn't have ceph_openat(), so for now assert this.
 414          */
 415         SMB_ASSERT(dirfsp->fh->fd == AT_FDCWD);
 416
 417         DBG_DEBUG("[CEPH] openat(%p, %s, %p, %d, %d)\n", handle,
 418                   smb_fname_str_dbg(smb_fname), fsp, flags, mode);
 419
 420         if (smb_fname->stream_name) {
 421                 goto out;
 422         }
 423
 424         result = ceph_open(handle->data, smb_fname->base_name, flags, mode);
 425 out:
 426         DBG_DEBUG("[CEPH] open(...) = %d\n", result);
 427         WRAP_RETURN(result);
 428 }
 429
 430 static int cephwrap_close(struct vfs_handle_struct *handle, files_struct *fsp)
 431 {
 432         int result;
 433
 434         DBG_DEBUG("[CEPH] close(%p, %p)\n", handle, fsp);
 435         result = ceph_close(handle->data, fsp->fh->fd);
 436         DBG_DEBUG("[CEPH] close(...) = %d\n", result);
 437
 438         WRAP_RETURN(result);
 439 }
 440
 441 static ssize_t cephwrap_pread(struct vfs_handle_struct *handle, files_struct *fsp, void *data,
 442                         size_t n, off_t offset)
 443 {
 444         ssize_t result;
 445
 446         DBG_DEBUG("[CEPH] pread(%p, %p, %p, %llu, %llu)\n", handle, fsp, data, llu(n), llu(offset));
 447
 448         result = ceph_read(handle->data, fsp->fh->fd, data, n, offset);
 449         DBG_DEBUG("[CEPH] pread(...) = %llu\n", llu(result));
 450         WRAP_RETURN(result);
 451 }
 452
 453 struct cephwrap_pread_state {
 454         ssize_t bytes_read;
 455         struct vfs_aio_state vfs_aio_state;
 456 };
 457
 458 /*
 459  * Fake up an async ceph read by calling the synchronous API.
 460  */
 461 static struct tevent_req *cephwrap_pread_send(struct vfs_handle_struct *handle,
 462                                               TALLOC_CTX *mem_ctx,
 463                                               struct tevent_context *ev,
 464                                               struct files_struct *fsp,
 465                                               void *data,
 466                                               size_t n, off_t offset)
 467 {
 468         struct tevent_req *req = NULL;
 469         struct cephwrap_pread_state *state = NULL;
 470         int ret = -1;
 471
 472         DBG_DEBUG("[CEPH] %s\n", __func__);
 473         req = tevent_req_create(mem_ctx, &state, struct cephwrap_pread_state);
 474         if (req == NULL) {
 475                 return NULL;
 476         }
 477
 478         ret = ceph_read(handle->data, fsp->fh->fd, data, n, offset);
 479         if (ret < 0) {
 480                 /* ceph returns -errno on error. */
 481                 tevent_req_error(req, -ret);
 482                 return tevent_req_post(req, ev);
 483         }
 484
 485         state->bytes_read = ret;
 486         tevent_req_done(req);
 487         /* Return and schedule the completion of the call. */
 488         return tevent_req_post(req, ev);
 489 }
 490
 491 static ssize_t cephwrap_pread_recv(struct tevent_req *req,
 492                                    struct vfs_aio_state *vfs_aio_state)
 493 {
 494         struct cephwrap_pread_state *state =
 495                 tevent_req_data(req, struct cephwrap_pread_state);
 496
 497         DBG_DEBUG("[CEPH] %s\n", __func__);
 498         if (tevent_req_is_unix_error(req, &vfs_aio_state->error)) {
 499                 return -1;
 500         }
 501         *vfs_aio_state = state->vfs_aio_state;
 502         return state->bytes_read;
 503 }
 504
 505 static ssize_t cephwrap_pwrite(struct vfs_handle_struct *handle, files_struct *fsp, const void *data,
 506                         size_t n, off_t offset)
 507 {
 508         ssize_t result;
 509
 510         DBG_DEBUG("[CEPH] pwrite(%p, %p, %p, %llu, %llu)\n", handle, fsp, data, llu(n), llu(offset));
 511         result = ceph_write(handle->data, fsp->fh->fd, data, n, offset);
 512         DBG_DEBUG("[CEPH] pwrite(...) = %llu\n", llu(result));
 513         WRAP_RETURN(result);
 514 }
 515
 516 struct cephwrap_pwrite_state {
 517         ssize_t bytes_written;
 518         struct vfs_aio_state vfs_aio_state;
 519 };
 520
 521 /*
 522  * Fake up an async ceph write by calling the synchronous API.
 523  */
 524 static struct tevent_req *cephwrap_pwrite_send(struct vfs_handle_struct *handle,
 525                                                TALLOC_CTX *mem_ctx,
 526                                                struct tevent_context *ev,
 527                                                struct files_struct *fsp,
 528                                                const void *data,
 529                                                size_t n, off_t offset)
 530 {
 531         struct tevent_req *req = NULL;
 532         struct cephwrap_pwrite_state *state = NULL;
 533         int ret = -1;
 534
 535         DBG_DEBUG("[CEPH] %s\n", __func__);
 536         req = tevent_req_create(mem_ctx, &state, struct cephwrap_pwrite_state);
 537         if (req == NULL) {
 538                 return NULL;
 539         }
 540
 541         ret = ceph_write(handle->data, fsp->fh->fd, data, n, offset);
 542         if (ret < 0) {
 543                 /* ceph returns -errno on error. */
 544                 tevent_req_error(req, -ret);
 545                 return tevent_req_post(req, ev);
 546         }
 547
 548         state->bytes_written = ret;
 549         tevent_req_done(req);
 550         /* Return and schedule the completion of the call. */
 551         return tevent_req_post(req, ev);
 552 }
 553
 554 static ssize_t cephwrap_pwrite_recv(struct tevent_req *req,
 555                                     struct vfs_aio_state *vfs_aio_state)
 556 {
 557         struct cephwrap_pwrite_state *state =
 558                 tevent_req_data(req, struct cephwrap_pwrite_state);
 559
 560         DBG_DEBUG("[CEPH] %s\n", __func__);
 561         if (tevent_req_is_unix_error(req, &vfs_aio_state->error)) {
 562                 return -1;
 563         }
 564         *vfs_aio_state = state->vfs_aio_state;
 565         return state->bytes_written;
 566 }
 567
 568 static off_t cephwrap_lseek(struct vfs_handle_struct *handle, files_struct *fsp, off_t offset, int whence)
 569 {
 570         off_t result = 0;
 571
 572         DBG_DEBUG("[CEPH] cephwrap_lseek\n");
 573         result = ceph_lseek(handle->data, fsp->fh->fd, offset, whence);
 574         WRAP_RETURN(result);
 575 }
 576
 577 static ssize_t cephwrap_sendfile(struct vfs_handle_struct *handle, int tofd, files_struct *fromfsp, const DATA_BLOB *hdr,
 578                         off_t offset, size_t n)
 579 {
 580         /*
 581          * We cannot support sendfile because libceph is in user space.
 582          */
 583         DBG_DEBUG("[CEPH] cephwrap_sendfile\n");
 584         errno = ENOTSUP;
 585         return -1;
 586 }
 587
 588 static ssize_t cephwrap_recvfile(struct vfs_handle_struct *handle,
 589                         int fromfd,
 590                         files_struct *tofsp,
 591                         off_t offset,
 592                         size_t n)
 593 {
 594         /*
 595          * We cannot support recvfile because libceph is in user space.
 596          */
 597         DBG_DEBUG("[CEPH] cephwrap_recvfile\n");
 598         errno=ENOTSUP;
 599         return -1;
 600 }
 601
 602 static int cephwrap_renameat(struct vfs_handle_struct *handle,
 603                         files_struct *srcfsp,
 604                         const struct smb_filename *smb_fname_src,
 605                         files_struct *dstfsp,
 606                         const struct smb_filename *smb_fname_dst)
 607 {
 608         int result = -1;
 609         DBG_DEBUG("[CEPH] cephwrap_renameat\n");
 610         if (smb_fname_src->stream_name || smb_fname_dst->stream_name) {
 611                 errno = ENOENT;
 612                 return result;
 613         }
 614
 615         SMB_ASSERT(srcfsp == srcfsp->conn->cwd_fsp);
 616         SMB_ASSERT(dstfsp == dstfsp->conn->cwd_fsp);
 617
 618         result = ceph_rename(handle->data, smb_fname_src->base_name, smb_fname_dst->base_name);
 619         WRAP_RETURN(result);
 620 }
 621
 622 /*
 623  * Fake up an async ceph fsync by calling the synchronous API.
 624  */
 625
 626 static struct tevent_req *cephwrap_fsync_send(struct vfs_handle_struct *handle,
 627                                         TALLOC_CTX *mem_ctx,
 628                                         struct tevent_context *ev,
 629                                         files_struct *fsp)
 630 {
 631         struct tevent_req *req = NULL;
 632         struct vfs_aio_state *state = NULL;
 633         int ret = -1;
 634
 635         DBG_DEBUG("[CEPH] cephwrap_fsync_send\n");
 636
 637         req = tevent_req_create(mem_ctx, &state, struct vfs_aio_state);
 638         if (req == NULL) {
 639                 return NULL;
 640         }
 641
 642         /* Make sync call. */
 643         ret = ceph_fsync(handle->data, fsp->fh->fd, false);
 644
 645         if (ret != 0) {
 646                 /* ceph_fsync returns -errno on error. */
 647                 tevent_req_error(req, -ret);
 648                 return tevent_req_post(req, ev);
 649         }
 650
 651         /* Mark it as done. */
 652         tevent_req_done(req);
 653         /* Return and schedule the completion of the call. */
 654         return tevent_req_post(req, ev);
 655 }
 656
 657 static int cephwrap_fsync_recv(struct tevent_req *req,
 658                                 struct vfs_aio_state *vfs_aio_state)
 659 {
 660         struct vfs_aio_state *state =
 661                 tevent_req_data(req, struct vfs_aio_state);
 662
 663         DBG_DEBUG("[CEPH] cephwrap_fsync_recv\n");
 664
 665         if (tevent_req_is_unix_error(req, &vfs_aio_state->error)) {
 666                 return -1;
 667         }
 668         *vfs_aio_state = *state;
 669         return 0;
 670 }
 671
 672 #define SAMBA_STATX_ATTR_MASK   (CEPH_STATX_BASIC_STATS|CEPH_STATX_BTIME)
 673
 674 static void init_stat_ex_from_ceph_statx(struct stat_ex *dst, const struct ceph_statx *stx)
 675 {
 676         DBG_DEBUG("[CEPH]\tstx = {dev = %llx, ino = %llu, mode = 0x%x, "
 677                   "nlink = %llu, uid = %d, gid = %d, rdev = %llx, size = %llu, "
 678                   "blksize = %llu, blocks = %llu, atime = %llu, mtime = %llu, "
 679                   "ctime = %llu, btime = %llu}\n",
 680                   llu(stx->stx_dev), llu(stx->stx_ino), stx->stx_mode,
 681                   llu(stx->stx_nlink), stx->stx_uid, stx->stx_gid,
 682                   llu(stx->stx_rdev), llu(stx->stx_size), llu(stx->stx_blksize),
 683                   llu(stx->stx_blocks), llu(stx->stx_atime.tv_sec),
 684                   llu(stx->stx_mtime.tv_sec), llu(stx->stx_ctime.tv_sec),
 685                   llu(stx->stx_btime.tv_sec));
 686
 687         if ((stx->stx_mask & SAMBA_STATX_ATTR_MASK) != SAMBA_STATX_ATTR_MASK) {
 688                 DBG_WARNING("%s: stx->stx_mask is incorrect (wanted %x, got %x)",
 689                                 __func__, SAMBA_STATX_ATTR_MASK, stx->stx_mask);
 690         }
 691
 692         dst->st_ex_dev = stx->stx_dev;
 693         dst->st_ex_rdev = stx->stx_rdev;
 694         dst->st_ex_ino = stx->stx_ino;
 695         dst->st_ex_mode = stx->stx_mode;
 696         dst->st_ex_uid = stx->stx_uid;
 697         dst->st_ex_gid = stx->stx_gid;
 698         dst->st_ex_size = stx->stx_size;
 699         dst->st_ex_nlink = stx->stx_nlink;
 700         dst->st_ex_atime = stx->stx_atime;
 701         dst->st_ex_btime = stx->stx_btime;
 702         dst->st_ex_ctime = stx->stx_ctime;
 703         dst->st_ex_mtime = stx->stx_mtime;
 704         dst->st_ex_itime = dst->st_ex_btime;
 705         dst->st_ex_iflags = ST_EX_IFLAG_CALCULATED_ITIME;
 706         dst->st_ex_blksize = stx->stx_blksize;
 707         dst->st_ex_blocks = stx->stx_blocks;
 708         dst->st_ex_file_id = dst->st_ex_ino;
 709         dst->st_ex_iflags |= ST_EX_IFLAG_CALCULATED_FILE_ID;
 710 }
 711
 712 static int cephwrap_stat(struct vfs_handle_struct *handle,
 713                         struct smb_filename *smb_fname)
 714 {
 715         int result = -1;
 716         struct ceph_statx stx;
 717
 718         DBG_DEBUG("[CEPH] stat(%p, %s)\n", handle, smb_fname_str_dbg(smb_fname));
 719
 720         if (smb_fname->stream_name) {
 721                 errno = ENOENT;
 722                 return result;
 723         }
 724
 725         result = ceph_statx(handle->data, smb_fname->base_name, &stx,
 726                                 SAMBA_STATX_ATTR_MASK, 0);
 727         DBG_DEBUG("[CEPH] statx(...) = %d\n", result);
 728         if (result < 0) {
 729                 WRAP_RETURN(result);
 730         }
 731
 732         init_stat_ex_from_ceph_statx(&smb_fname->st, &stx);
 733         DBG_DEBUG("[CEPH] mode = 0x%x\n", smb_fname->st.st_ex_mode);
 734         return result;
 735 }
 736
 737 static int cephwrap_fstat(struct vfs_handle_struct *handle, files_struct *fsp, SMB_STRUCT_STAT *sbuf)
 738 {
 739         int result = -1;
 740         struct ceph_statx stx;
 741
 742         DBG_DEBUG("[CEPH] fstat(%p, %d)\n", handle, fsp->fh->fd);
 743         result = ceph_fstatx(handle->data, fsp->fh->fd, &stx,
 744                                 SAMBA_STATX_ATTR_MASK, 0);
 745         DBG_DEBUG("[CEPH] fstat(...) = %d\n", result);
 746         if (result < 0) {
 747                 WRAP_RETURN(result);
 748         }
 749
 750         init_stat_ex_from_ceph_statx(sbuf, &stx);
 751         DBG_DEBUG("[CEPH] mode = 0x%x\n", sbuf->st_ex_mode);
 752         return result;
 753 }
 754
 755 static int cephwrap_lstat(struct vfs_handle_struct *handle,
 756                          struct smb_filename *smb_fname)
 757 {
 758         int result = -1;
 759         struct ceph_statx stx;
 760
 761         DBG_DEBUG("[CEPH] lstat(%p, %s)\n", handle, smb_fname_str_dbg(smb_fname));
 762
 763         if (smb_fname->stream_name) {
 764                 errno = ENOENT;
 765                 return result;
 766         }
 767
 768         result = ceph_statx(handle->data, smb_fname->base_name, &stx,
 769                                 SAMBA_STATX_ATTR_MASK, AT_SYMLINK_NOFOLLOW);
 770         DBG_DEBUG("[CEPH] lstat(...) = %d\n", result);
 771         if (result < 0) {
 772                 WRAP_RETURN(result);
 773         }
 774
 775         init_stat_ex_from_ceph_statx(&smb_fname->st, &stx);
 776         return result;
 777 }
 778
 779 static int cephwrap_ntimes(struct vfs_handle_struct *handle,
 780                          const struct smb_filename *smb_fname,
 781                          struct smb_file_time *ft)
 782 {
 783         struct ceph_statx stx = { 0 };
 784         int result;
 785         int mask = 0;
 786
 787         if (!is_omit_timespec(&ft->atime)) {
 788                 stx.stx_atime = ft->atime;
 789                 mask |= CEPH_SETATTR_ATIME;
 790         }
 791         if (!is_omit_timespec(&ft->mtime)) {
 792                 stx.stx_mtime = ft->mtime;
 793                 mask |= CEPH_SETATTR_MTIME;
 794         }
 795         if (!is_omit_timespec(&ft->create_time)) {
 796                 stx.stx_btime = ft->create_time;
 797                 mask |= CEPH_SETATTR_BTIME;
 798         }
 799
 800         if (!mask) {
 801                 return 0;
 802         }
 803
 804         result = ceph_setattrx(handle->data, smb_fname->base_name, &stx, mask, 0);
 805         DBG_DEBUG("[CEPH] ntimes(%p, %s, {%ld, %ld, %ld, %ld}) = %d\n", handle, smb_fname_str_dbg(smb_fname),
 806                                 ft->mtime.tv_sec, ft->atime.tv_sec, ft->ctime.tv_sec,
 807                                 ft->create_time.tv_sec, result);
 808         return result;
 809 }
 810
 811 static int cephwrap_unlinkat(struct vfs_handle_struct *handle,
 812                         struct files_struct *dirfsp,
 813                         const struct smb_filename *smb_fname,
 814                         int flags)
 815 {
 816         int result = -1;
 817
 818         DBG_DEBUG("[CEPH] unlink(%p, %s)\n",
 819                 handle,
 820                 smb_fname_str_dbg(smb_fname));
 821         SMB_ASSERT(dirfsp == dirfsp->conn->cwd_fsp);
 822         if (smb_fname->stream_name) {
 823                 errno = ENOENT;
 824                 return result;
 825         }
 826         if (flags & AT_REMOVEDIR) {
 827                 result = ceph_rmdir(handle->data, smb_fname->base_name);
 828         } else {
 829                 result = ceph_unlink(handle->data, smb_fname->base_name);
 830         }
 831         DBG_DEBUG("[CEPH] unlink(...) = %d\n", result);
 832         WRAP_RETURN(result);
 833 }
 834
 835 static int cephwrap_chmod(struct vfs_handle_struct *handle,
 836                         const struct smb_filename *smb_fname,
 837                         mode_t mode)
 838 {
 839         int result;
 840
 841         DBG_DEBUG("[CEPH] chmod(%p, %s, %d)\n", handle, smb_fname->base_name, mode);
 842         result = ceph_chmod(handle->data, smb_fname->base_name, mode);
 843         DBG_DEBUG("[CEPH] chmod(...) = %d\n", result);
 844         WRAP_RETURN(result);
 845 }
 846
 847 static int cephwrap_fchmod(struct vfs_handle_struct *handle, files_struct *fsp, mode_t mode)
 848 {
 849         int result;
 850
 851         DBG_DEBUG("[CEPH] fchmod(%p, %p, %d)\n", handle, fsp, mode);
 852         result = ceph_fchmod(handle->data, fsp->fh->fd, mode);
 853         DBG_DEBUG("[CEPH] fchmod(...) = %d\n", result);
 854         WRAP_RETURN(result);
 855 }
 856
 857 static int cephwrap_fchown(struct vfs_handle_struct *handle, files_struct *fsp, uid_t uid, gid_t gid)
 858 {
 859         int result;
 860
 861         DBG_DEBUG("[CEPH] fchown(%p, %p, %d, %d)\n", handle, fsp, uid, gid);
 862         result = ceph_fchown(handle->data, fsp->fh->fd, uid, gid);
 863         DBG_DEBUG("[CEPH] fchown(...) = %d\n", result);
 864         WRAP_RETURN(result);
 865 }
 866
 867 static int cephwrap_lchown(struct vfs_handle_struct *handle,
 868                         const struct smb_filename *smb_fname,
 869                         uid_t uid,
 870                         gid_t gid)
 871 {
 872         int result;
 873         DBG_DEBUG("[CEPH] lchown(%p, %s, %d, %d)\n", handle, smb_fname->base_name, uid, gid);
 874         result = ceph_lchown(handle->data, smb_fname->base_name, uid, gid);
 875         DBG_DEBUG("[CEPH] lchown(...) = %d\n", result);
 876         WRAP_RETURN(result);
 877 }
 878
 879 static int cephwrap_chdir(struct vfs_handle_struct *handle,
 880                         const struct smb_filename *smb_fname)
 881 {
 882         int result = -1;
 883         DBG_DEBUG("[CEPH] chdir(%p, %s)\n", handle, smb_fname->base_name);
 884         result = ceph_chdir(handle->data, smb_fname->base_name);
 885         DBG_DEBUG("[CEPH] chdir(...) = %d\n", result);
 886         WRAP_RETURN(result);
 887 }
 888
 889 static struct smb_filename *cephwrap_getwd(struct vfs_handle_struct *handle,
 890                         TALLOC_CTX *ctx)
 891 {
 892         const char *cwd = ceph_getcwd(handle->data);
 893         DBG_DEBUG("[CEPH] getwd(%p) = %s\n", handle, cwd);
 894         return synthetic_smb_fname(ctx,
 895                                 cwd,
 896                                 NULL,
 897                                 NULL,
 898                                 0,
 899                                 0);
 900 }
 901
 902 static int strict_allocate_ftruncate(struct vfs_handle_struct *handle, files_struct *fsp, off_t len)
 903 {
 904         off_t space_to_write;
 905         int result;
 906         NTSTATUS status;
 907         SMB_STRUCT_STAT *pst;
 908
 909         status = vfs_stat_fsp(fsp);
 910         if (!NT_STATUS_IS_OK(status)) {
 911                 return -1;
 912         }
 913         pst = &fsp->fsp_name->st;
 914
 915 #ifdef S_ISFIFO
 916         if (S_ISFIFO(pst->st_ex_mode))
 917                 return 0;
 918 #endif
 919
 920         if (pst->st_ex_size == len)
 921                 return 0;
 922
 923         /* Shrink - just ftruncate. */
 924         if (pst->st_ex_size > len) {
 925                 result = ceph_ftruncate(handle->data, fsp->fh->fd, len);
 926                 WRAP_RETURN(result);
 927         }
 928
 929         space_to_write = len - pst->st_ex_size;
 930         result = ceph_fallocate(handle->data, fsp->fh->fd, 0, pst->st_ex_size,
 931                                 space_to_write);
 932         WRAP_RETURN(result);
 933 }
 934
 935 static int cephwrap_ftruncate(struct vfs_handle_struct *handle, files_struct *fsp, off_t len)
 936 {
 937         int result = -1;
 938
 939         DBG_DEBUG("[CEPH] ftruncate(%p, %p, %llu\n", handle, fsp, llu(len));
 940
 941         if (lp_strict_allocate(SNUM(fsp->conn))) {
 942                 return strict_allocate_ftruncate(handle, fsp, len);
 943         }
 944
 945         result = ceph_ftruncate(handle->data, fsp->fh->fd, len);
 946         WRAP_RETURN(result);
 947 }
 948
 949 static int cephwrap_fallocate(struct vfs_handle_struct *handle,
 950                               struct files_struct *fsp,
 951                               uint32_t mode,
 952                               off_t offset,
 953                               off_t len)
 954 {
 955         int result;
 956
 957         DBG_DEBUG("[CEPH] fallocate(%p, %p, %u, %llu, %llu\n",
 958                   handle, fsp, mode, llu(offset), llu(len));
 959         /* unsupported mode flags are rejected by libcephfs */
 960         result = ceph_fallocate(handle->data, fsp->fh->fd, mode, offset, len);
 961         DBG_DEBUG("[CEPH] fallocate(...) = %d\n", result);
 962         WRAP_RETURN(result);
 963 }
 964
 965 static bool cephwrap_lock(struct vfs_handle_struct *handle, files_struct *fsp, int op, off_t offset, off_t count, int type)
 966 {
 967         DBG_DEBUG("[CEPH] lock\n");
 968         return true;
 969 }
 970
 971 static int cephwrap_kernel_flock(struct vfs_handle_struct *handle,
 972                                  files_struct *fsp,
 973                                  uint32_t share_access,
 974                                  uint32_t access_mask)
 975 {
 976         DBG_ERR("[CEPH] flock unsupported! Consider setting "
 977                 "\"kernel share modes = no\"\n");
 978
 979         errno = ENOSYS;
 980         return -1;
 981 }
 982
 983 static int cephwrap_fcntl(vfs_handle_struct *handle,
 984                           files_struct *fsp, int cmd, va_list cmd_arg)
 985 {
 986         /*
 987          * SMB_VFS_FCNTL() is currently only called by vfs_set_blocking() to
 988          * clear O_NONBLOCK, etc for LOCK_MAND and FIFOs. Ignore it.
 989          */
 990         if (cmd == F_GETFL) {
 991                 return 0;
 992         } else if (cmd == F_SETFL) {
 993                 va_list dup_cmd_arg;
 994                 int opt;
 995
 996                 va_copy(dup_cmd_arg, cmd_arg);
 997                 opt = va_arg(dup_cmd_arg, int);
 998                 va_end(dup_cmd_arg);
 999                 if (opt == 0) {
1000                         return 0;
1001                 }
1002                 DBG_ERR("unexpected fcntl SETFL(%d)\n", opt);
1003                 goto err_out;
1004         }
1005         DBG_ERR("unexpected fcntl: %d\n", cmd);
1006 err_out:
1007         errno = EINVAL;
1008         return -1;
1009 }
1010
1011 static bool cephwrap_getlock(struct vfs_handle_struct *handle, files_struct *fsp, off_t *poffset, off_t *pcount, int *ptype, pid_t *ppid)
1012 {
1013         DBG_DEBUG("[CEPH] getlock returning false and errno=0\n");
1014
1015         errno = 0;
1016         return false;
1017 }
1018
1019 /*
1020  * We cannot let this fall through to the default, because the file might only
1021  * be accessible from libceph (which is a user-space client) but the fd might
1022  * be for some file the kernel knows about.
1023  */
1024 static int cephwrap_linux_setlease(struct vfs_handle_struct *handle, files_struct *fsp,
1025                                 int leasetype)
1026 {
1027         int result = -1;
1028
1029         DBG_DEBUG("[CEPH] linux_setlease\n");
1030         errno = ENOSYS;
1031         return result;
1032 }
1033
1034 static int cephwrap_symlinkat(struct vfs_handle_struct *handle,
1035                 const struct smb_filename *link_target,
1036                 struct files_struct *dirfsp,
1037                 const struct smb_filename *new_smb_fname)
1038 {
1039         int result = -1;
1040         DBG_DEBUG("[CEPH] symlink(%p, %s, %s)\n", handle,
1041                         link_target->base_name,
1042                         new_smb_fname->base_name);
1043
1044         SMB_ASSERT(dirfsp == dirfsp->conn->cwd_fsp);
1045
1046         result = ceph_symlink(handle->data,
1047                         link_target->base_name,
1048                         new_smb_fname->base_name);
1049         DBG_DEBUG("[CEPH] symlink(...) = %d\n", result);
1050         WRAP_RETURN(result);
1051 }
1052
1053 static int cephwrap_readlinkat(struct vfs_handle_struct *handle,
1054                 files_struct *dirfsp,
1055                 const struct smb_filename *smb_fname,
1056                 char *buf,
1057                 size_t bufsiz)
1058 {
1059         int result = -1;
1060         DBG_DEBUG("[CEPH] readlink(%p, %s, %p, %llu)\n", handle,
1061                         smb_fname->base_name, buf, llu(bufsiz));
1062
1063         SMB_ASSERT(dirfsp == dirfsp->conn->cwd_fsp);
1064
1065         result = ceph_readlink(handle->data, smb_fname->base_name, buf, bufsiz);
1066         DBG_DEBUG("[CEPH] readlink(...) = %d\n", result);
1067         WRAP_RETURN(result);
1068 }
1069
1070 static int cephwrap_linkat(struct vfs_handle_struct *handle,
1071                 files_struct *srcfsp,
1072                 const struct smb_filename *old_smb_fname,
1073                 files_struct *dstfsp,
1074                 const struct smb_filename *new_smb_fname,
1075                 int flags)
1076 {
1077         int result = -1;
1078         DBG_DEBUG("[CEPH] link(%p, %s, %s)\n", handle,
1079                         old_smb_fname->base_name,
1080                         new_smb_fname->base_name);
1081
1082         SMB_ASSERT(srcfsp == srcfsp->conn->cwd_fsp);
1083         SMB_ASSERT(dstfsp == dstfsp->conn->cwd_fsp);
1084
1085         result = ceph_link(handle->data,
1086                                 old_smb_fname->base_name,
1087                                 new_smb_fname->base_name);
1088         DBG_DEBUG("[CEPH] link(...) = %d\n", result);
1089         WRAP_RETURN(result);
1090 }
1091
1092 static int cephwrap_mknodat(struct vfs_handle_struct *handle,
1093                 files_struct *dirfsp,
1094                 const struct smb_filename *smb_fname,
1095                 mode_t mode,
1096                 SMB_DEV_T dev)
1097 {
1098         int result = -1;
1099         DBG_DEBUG("[CEPH] mknodat(%p, %s)\n", handle, smb_fname->base_name);
1100         SMB_ASSERT(dirfsp == dirfsp->conn->cwd_fsp);
1101         result = ceph_mknod(handle->data, smb_fname->base_name, mode, dev);
1102         DBG_DEBUG("[CEPH] mknodat(...) = %d\n", result);
1103         WRAP_RETURN(result);
1104 }
1105
1106 /*
1107  * This is a simple version of real-path ... a better version is needed to
1108  * ask libceph about symbolic links.
1109  */
1110 static struct smb_filename *cephwrap_realpath(struct vfs_handle_struct *handle,
1111                                 TALLOC_CTX *ctx,
1112                                 const struct smb_filename *smb_fname)
1113 {
1114         char *result = NULL;
1115         const char *path = smb_fname->base_name;
1116         size_t len = strlen(path);
1117         struct smb_filename *result_fname = NULL;
1118         int r = -1;
1119
1120         if (len && (path[0] == '/')) {
1121                 r = asprintf(&result, "%s", path);
1122         } else if ((len >= 2) && (path[0] == '.') && (path[1] == '/')) {
1123                 if (len == 2) {
1124                         r = asprintf(&result, "%s",
1125                                         handle->conn->cwd_fsp->fsp_name->base_name);
1126                 } else {
1127                         r = asprintf(&result, "%s/%s",
1128                                         handle->conn->cwd_fsp->fsp_name->base_name, &path[2]);
1129                 }
1130         } else {
1131                 r = asprintf(&result, "%s/%s",
1132                                 handle->conn->cwd_fsp->fsp_name->base_name, path);
1133         }
1134
1135         if (r < 0) {
1136                 return NULL;
1137         }
1138
1139         DBG_DEBUG("[CEPH] realpath(%p, %s) = %s\n", handle, path, result);
1140         result_fname = synthetic_smb_fname(ctx,
1141                                 result,
1142                                 NULL,
1143                                 NULL,
1144                                 0,
1145                                 0);
1146         SAFE_FREE(result);
1147         return result_fname;
1148 }
1149
1150 static int cephwrap_chflags(struct vfs_handle_struct *handle,
1151                         const struct smb_filename *smb_fname,
1152                         unsigned int flags)
1153 {
1154         errno = ENOSYS;
1155         return -1;
1156 }
1157
1158 static int cephwrap_get_real_filename(struct vfs_handle_struct *handle,
1159                                      const struct smb_filename *path,
1160                                      const char *name,
1161                                      TALLOC_CTX *mem_ctx,
1162                                      char **found_name)
1163 {
1164         /*
1165          * Don't fall back to get_real_filename so callers can differentiate
1166          * between a full directory scan and an actual case-insensitive stat.
1167          */
1168         errno = EOPNOTSUPP;
1169         return -1;
1170 }
1171
1172 static const char *cephwrap_connectpath(struct vfs_handle_struct *handle,
1173                                        const struct smb_filename *smb_fname)
1174 {
1175         return handle->conn->connectpath;
1176 }
1177
1178 /****************************************************************
1179  Extended attribute operations.
1180 *****************************************************************/
1181
1182 static ssize_t cephwrap_getxattr(struct vfs_handle_struct *handle,
1183                         const struct smb_filename *smb_fname,
1184                         const char *name,
1185                         void *value,
1186                         size_t size)
1187 {
1188         int ret;
1189         DBG_DEBUG("[CEPH] getxattr(%p, %s, %s, %p, %llu)\n", handle,
1190                         smb_fname->base_name, name, value, llu(size));
1191         ret = ceph_getxattr(handle->data,
1192                         smb_fname->base_name, name, value, size);
1193         DBG_DEBUG("[CEPH] getxattr(...) = %d\n", ret);
1194         if (ret < 0) {
1195                 WRAP_RETURN(ret);
1196         }
1197         return (ssize_t)ret;
1198 }
1199
1200 static ssize_t cephwrap_fgetxattr(struct vfs_handle_struct *handle, struct files_struct *fsp, const char *name, void *value, size_t size)
1201 {
1202         int ret;
1203         DBG_DEBUG("[CEPH] fgetxattr(%p, %p, %s, %p, %llu)\n", handle, fsp, name, value, llu(size));
1204         ret = ceph_fgetxattr(handle->data, fsp->fh->fd, name, value, size);
1205         DBG_DEBUG("[CEPH] fgetxattr(...) = %d\n", ret);
1206         if (ret < 0) {
1207                 WRAP_RETURN(ret);
1208         }
1209         return (ssize_t)ret;
1210 }
1211
1212 static ssize_t cephwrap_listxattr(struct vfs_handle_struct *handle,
1213                         const struct smb_filename *smb_fname,
1214                         char *list,
1215                         size_t size)
1216 {
1217         int ret;
1218         DBG_DEBUG("[CEPH] listxattr(%p, %s, %p, %llu)\n", handle,
1219                         smb_fname->base_name, list, llu(size));
1220         ret = ceph_listxattr(handle->data, smb_fname->base_name, list, size);
1221         DBG_DEBUG("[CEPH] listxattr(...) = %d\n", ret);
1222         if (ret < 0) {
1223                 WRAP_RETURN(ret);
1224         }
1225         return (ssize_t)ret;
1226 }
1227
1228 static ssize_t cephwrap_flistxattr(struct vfs_handle_struct *handle, struct files_struct *fsp, char *list, size_t size)
1229 {
1230         int ret;
1231         DBG_DEBUG("[CEPH] flistxattr(%p, %p, %p, %llu)\n",
1232                   handle, fsp, list, llu(size));
1233         ret = ceph_flistxattr(handle->data, fsp->fh->fd, list, size);
1234         DBG_DEBUG("[CEPH] flistxattr(...) = %d\n", ret);
1235         if (ret < 0) {
1236                 WRAP_RETURN(ret);
1237         }
1238         return (ssize_t)ret;
1239 }
1240
1241 static int cephwrap_removexattr(struct vfs_handle_struct *handle,
1242                                 const struct smb_filename *smb_fname,
1243                                 const char *name)
1244 {
1245         int ret;
1246         DBG_DEBUG("[CEPH] removexattr(%p, %s, %s)\n", handle,
1247                         smb_fname->base_name, name);
1248         ret = ceph_removexattr(handle->data, smb_fname->base_name, name);
1249         DBG_DEBUG("[CEPH] removexattr(...) = %d\n", ret);
1250         WRAP_RETURN(ret);
1251 }
1252
1253 static int cephwrap_fremovexattr(struct vfs_handle_struct *handle, struct files_struct *fsp, const char *name)
1254 {
1255         int ret;
1256         DBG_DEBUG("[CEPH] fremovexattr(%p, %p, %s)\n", handle, fsp, name);
1257         ret = ceph_fremovexattr(handle->data, fsp->fh->fd, name);
1258         DBG_DEBUG("[CEPH] fremovexattr(...) = %d\n", ret);
1259         WRAP_RETURN(ret);
1260 }
1261
1262 static int cephwrap_setxattr(struct vfs_handle_struct *handle,
1263                                 const struct smb_filename *smb_fname,
1264                                 const char *name,
1265                                 const void *value,
1266                                 size_t size,
1267                                 int flags)
1268 {
1269         int ret;
1270         DBG_DEBUG("[CEPH] setxattr(%p, %s, %s, %p, %llu, %d)\n", handle,
1271                         smb_fname->base_name, name, value, llu(size), flags);
1272         ret = ceph_setxattr(handle->data, smb_fname->base_name,
1273                         name, value, size, flags);
1274         DBG_DEBUG("[CEPH] setxattr(...) = %d\n", ret);
1275         WRAP_RETURN(ret);
1276 }
1277
1278 static int cephwrap_fsetxattr(struct vfs_handle_struct *handle, struct files_struct *fsp, const char *name, const void *value, size_t size, int flags)
1279 {
1280         int ret;
1281         DBG_DEBUG("[CEPH] fsetxattr(%p, %p, %s, %p, %llu, %d)\n", handle, fsp, name, value, llu(size), flags);
1282         ret = ceph_fsetxattr(handle->data, fsp->fh->fd,
1283                              name, value, size, flags);
1284         DBG_DEBUG("[CEPH] fsetxattr(...) = %d\n", ret);
1285         WRAP_RETURN(ret);
1286 }
1287
1288 static bool cephwrap_aio_force(struct vfs_handle_struct *handle, struct files_struct *fsp)
1289 {
1290
1291         /*
1292          * We do not support AIO yet.
1293          */
1294
1295         DBG_DEBUG("[CEPH] cephwrap_aio_force(%p, %p) = false (errno = ENOTSUP)\n", handle, fsp);
1296         errno = ENOTSUP;
1297         return false;
1298 }
1299
1300 static NTSTATUS cephwrap_create_dfs_pathat(struct vfs_handle_struct *handle,
1301                                 struct files_struct *dirfsp,
1302                                 const struct smb_filename *smb_fname,
1303                                 const struct referral *reflist,
1304                                 size_t referral_count)
1305 {
1306         TALLOC_CTX *frame = talloc_stackframe();
1307         NTSTATUS status = NT_STATUS_NO_MEMORY;
1308         int ret;
1309         char *msdfs_link = NULL;
1310
1311         SMB_ASSERT(dirfsp == dirfsp->conn->cwd_fsp);
1312
1313         /* Form the msdfs_link contents */
1314         msdfs_link = msdfs_link_string(frame,
1315                                         reflist,
1316                                         referral_count);
1317         if (msdfs_link == NULL) {
1318                 goto out;
1319         }
1320
1321         ret = ceph_symlink(handle->data,
1322                         msdfs_link,
1323                         smb_fname->base_name);
1324         if (ret == 0) {
1325                 status = NT_STATUS_OK;
1326         } else {
1327                 status = map_nt_error_from_unix(-ret);
1328         }
1329
1330   out:
1331
1332         DBG_DEBUG("[CEPH] create_dfs_pathat(%s) = %s\n",
1333                         smb_fname->base_name,
1334                         nt_errstr(status));
1335
1336         TALLOC_FREE(frame);
1337         return status;
1338 }
1339
1340 /*
1341  * Read and return the contents of a DFS redirect given a
1342  * pathname. A caller can pass in NULL for ppreflist and
1343  * preferral_count but still determine if this was a
1344  * DFS redirect point by getting NT_STATUS_OK back
1345  * without incurring the overhead of reading and parsing
1346  * the referral contents.
1347  */
1348
1349 static NTSTATUS cephwrap_read_dfs_pathat(struct vfs_handle_struct *handle,
1350                                 TALLOC_CTX *mem_ctx,
1351                                 struct files_struct *dirfsp,
1352                                 struct smb_filename *smb_fname,
1353                                 struct referral **ppreflist,
1354                                 size_t *preferral_count)
1355 {
1356         NTSTATUS status = NT_STATUS_NO_MEMORY;
1357         size_t bufsize;
1358         char *link_target = NULL;
1359         int referral_len;
1360         bool ok;
1361 #if defined(HAVE_BROKEN_READLINK)
1362         char link_target_buf[PATH_MAX];
1363 #else
1364         char link_target_buf[7];
1365 #endif
1366         struct ceph_statx stx;
1367         int ret;
1368
1369         SMB_ASSERT(dirfsp == dirfsp->conn->cwd_fsp);
1370
1371         if (is_named_stream(smb_fname)) {
1372                 status = NT_STATUS_OBJECT_NAME_NOT_FOUND;
1373                 goto err;
1374         }
1375
1376         if (ppreflist == NULL && preferral_count == NULL) {
1377                 /*
1378                  * We're only checking if this is a DFS
1379                  * redirect. We don't need to return data.
1380                  */
1381                 bufsize = sizeof(link_target_buf);
1382                 link_target = link_target_buf;
1383         } else {
1384                 bufsize = PATH_MAX;
1385                 link_target = talloc_array(mem_ctx, char, bufsize);
1386                 if (!link_target) {
1387                         goto err;
1388                 }
1389         }
1390
1391         ret = ceph_statx(handle->data,
1392                          smb_fname->base_name,
1393                          &stx,
1394                          SAMBA_STATX_ATTR_MASK,
1395                          AT_SYMLINK_NOFOLLOW);
1396         if (ret < 0) {
1397                 status = map_nt_error_from_unix(-ret);
1398                 goto err;
1399         }
1400
1401         referral_len = ceph_readlink(handle->data,
1402                                 smb_fname->base_name,
1403                                 link_target,
1404                                 bufsize - 1);
1405         if (referral_len < 0) {
1406                 /* ceph errors are -errno. */
1407                 if (-referral_len == EINVAL) {
1408                         DBG_INFO("%s is not a link.\n",
1409                                 smb_fname->base_name);
1410                         status = NT_STATUS_OBJECT_TYPE_MISMATCH;
1411                 } else {
1412                         status = map_nt_error_from_unix(-referral_len);
1413                         DBG_ERR("Error reading "
1414                                 "msdfs link %s: %s\n",
1415                                 smb_fname->base_name,
1416                         strerror(errno));
1417                 }
1418                 goto err;
1419         }
1420         link_target[referral_len] = '\0';
1421
1422         DBG_INFO("%s -> %s\n",
1423                         smb_fname->base_name,
1424                         link_target);
1425
1426         if (!strnequal(link_target, "msdfs:", 6)) {
1427                 status = NT_STATUS_OBJECT_TYPE_MISMATCH;
1428                 goto err;
1429         }
1430
1431         if (ppreflist == NULL && preferral_count == NULL) {
1432                 /* Early return for checking if this is a DFS link. */
1433                 init_stat_ex_from_ceph_statx(&smb_fname->st, &stx);
1434                 return NT_STATUS_OK;
1435         }
1436
1437         ok = parse_msdfs_symlink(mem_ctx,
1438                         lp_msdfs_shuffle_referrals(SNUM(handle->conn)),
1439                         link_target,
1440                         ppreflist,
1441                         preferral_count);
1442
1443         if (ok) {
1444                 init_stat_ex_from_ceph_statx(&smb_fname->st, &stx);
1445                 status = NT_STATUS_OK;
1446         } else {
1447                 status = NT_STATUS_NO_MEMORY;
1448         }
1449
1450   err:
1451
1452         if (link_target != link_target_buf) {
1453                 TALLOC_FREE(link_target);
1454         }
1455         return status;
1456 }
1457
1458 static struct vfs_fn_pointers ceph_fns = {
1459         /* Disk operations */
1460
1461         .connect_fn = cephwrap_connect,
1462         .disconnect_fn = cephwrap_disconnect,
1463         .disk_free_fn = cephwrap_disk_free,
1464         .get_quota_fn = cephwrap_get_quota,
1465         .set_quota_fn = cephwrap_set_quota,
1466         .statvfs_fn = cephwrap_statvfs,
1467         .fs_capabilities_fn = cephwrap_fs_capabilities,
1468
1469         /* Directory operations */
1470
1471         .fdopendir_fn = cephwrap_fdopendir,
1472         .readdir_fn = cephwrap_readdir,
1473         .seekdir_fn = cephwrap_seekdir,
1474         .telldir_fn = cephwrap_telldir,
1475         .rewind_dir_fn = cephwrap_rewinddir,
1476         .mkdirat_fn = cephwrap_mkdirat,
1477         .closedir_fn = cephwrap_closedir,
1478
1479         /* File operations */
1480
1481         .create_dfs_pathat_fn = cephwrap_create_dfs_pathat,
1482         .read_dfs_pathat_fn = cephwrap_read_dfs_pathat,
1483         .openat_fn = cephwrap_openat,
1484         .close_fn = cephwrap_close,
1485         .pread_fn = cephwrap_pread,
1486         .pread_send_fn = cephwrap_pread_send,
1487         .pread_recv_fn = cephwrap_pread_recv,
1488         .pwrite_fn = cephwrap_pwrite,
1489         .pwrite_send_fn = cephwrap_pwrite_send,
1490         .pwrite_recv_fn = cephwrap_pwrite_recv,
1491         .lseek_fn = cephwrap_lseek,
1492         .sendfile_fn = cephwrap_sendfile,
1493         .recvfile_fn = cephwrap_recvfile,
1494         .renameat_fn = cephwrap_renameat,
1495         .fsync_send_fn = cephwrap_fsync_send,
1496         .fsync_recv_fn = cephwrap_fsync_recv,
1497         .stat_fn = cephwrap_stat,
1498         .fstat_fn = cephwrap_fstat,
1499         .lstat_fn = cephwrap_lstat,
1500         .unlinkat_fn = cephwrap_unlinkat,
1501         .chmod_fn = cephwrap_chmod,
1502         .fchmod_fn = cephwrap_fchmod,
1503         .fchown_fn = cephwrap_fchown,
1504         .lchown_fn = cephwrap_lchown,
1505         .chdir_fn = cephwrap_chdir,
1506         .getwd_fn = cephwrap_getwd,
1507         .ntimes_fn = cephwrap_ntimes,
1508         .ftruncate_fn = cephwrap_ftruncate,
1509         .fallocate_fn = cephwrap_fallocate,
1510         .lock_fn = cephwrap_lock,
1511         .kernel_flock_fn = cephwrap_kernel_flock,
1512         .fcntl_fn = cephwrap_fcntl,
1513         .linux_setlease_fn = cephwrap_linux_setlease,
1514         .getlock_fn = cephwrap_getlock,
1515         .symlinkat_fn = cephwrap_symlinkat,
1516         .readlinkat_fn = cephwrap_readlinkat,
1517         .linkat_fn = cephwrap_linkat,
1518         .mknodat_fn = cephwrap_mknodat,
1519         .realpath_fn = cephwrap_realpath,
1520         .chflags_fn = cephwrap_chflags,
1521         .get_real_filename_fn = cephwrap_get_real_filename,
1522         .connectpath_fn = cephwrap_connectpath,
1523
1524         /* EA operations. */
1525         .getxattr_fn = cephwrap_getxattr,
1526         .getxattrat_send_fn = vfs_not_implemented_getxattrat_send,
1527         .getxattrat_recv_fn = vfs_not_implemented_getxattrat_recv,
1528         .fgetxattr_fn = cephwrap_fgetxattr,
1529         .listxattr_fn = cephwrap_listxattr,
1530         .flistxattr_fn = cephwrap_flistxattr,
1531         .removexattr_fn = cephwrap_removexattr,
1532         .fremovexattr_fn = cephwrap_fremovexattr,
1533         .setxattr_fn = cephwrap_setxattr,
1534         .fsetxattr_fn = cephwrap_fsetxattr,
1535
1536         /* Posix ACL Operations */
1537         .sys_acl_get_file_fn = posixacl_xattr_acl_get_file,
1538         .sys_acl_get_fd_fn = posixacl_xattr_acl_get_fd,
1539         .sys_acl_blob_get_file_fn = posix_sys_acl_blob_get_file,
1540         .sys_acl_blob_get_fd_fn = posix_sys_acl_blob_get_fd,
1541         .sys_acl_set_file_fn = posixacl_xattr_acl_set_file,
1542         .sys_acl_set_fd_fn = posixacl_xattr_acl_set_fd,
1543         .sys_acl_delete_def_file_fn = posixacl_xattr_acl_delete_def_file,
1544
1545         /* aio operations */
1546         .aio_force_fn = cephwrap_aio_force,
1547 };
1548
1549 static_decl_vfs;
1550 NTSTATUS vfs_ceph_init(TALLOC_CTX *ctx)
1551 {
1552         return smb_register_vfs(SMB_VFS_INTERFACE_VERSION,
1553                                 "ceph", &ceph_fns);
1554 }